图片相似度比对

一、Web Worker方式实现 -- 浏览器

技术栈：
前端：vue2
比对算法：pixelmatch、resemble
worker与线程池：thread.js

Ⅰ. pixelmatch + thread.js

1. 依赖安装与配置

shell

# 安装threads node 和 浏览器可用
npm install threads tiny-worker
# 安装webpack插件。默认的worker.js 只支持es5，无法使用import 和 require。
npm install -D threads-plugin

vue.config.js

// vue.config.js
// 引入
const ThreadsPlugin = require('threads-plugin')
// 配置plugins, 支持es6写法，可以再worker中使用import

    configureWebpack: {
        resolve: {
            alias: {
                "@": resolve('src')
            }
        },
        plugins:[ new ThreadsPlugin()]
    }

2. 业务逻辑

如果计算量很大，我们不可能无限制的开线程，因此需要使用 队列 + 线程池的思想。
由此，我们使用Thread.js。 https://threads.js.org/

main线程

// 在处理业务逻辑的页面加入如下代码

// mock data
const leftImg = "https://ddiatmall.cafe24.com/web/product/medium/202103/1efbc777522e4279926b9acc02d4d1a7.jpg";
const eventData = {
   imgUrl: "http://proj.kr/china/uca/UCA-023/600_c/f4f4f4.jpg",
   listId: "1-PABSFWRWER0123",
 }
 const searchDataList = [];
 for (let i = 0; i < 10; i++) {
   searchDataList.push(eventData);
 }
 console.log("searchDataList ==", searchDataList)
 // 调用方法
 this.imgDiffTask(leftImg,searchDataList,0.1)


// 实际业务方法
async imgDiffTask(leftImg, searchImgList, threshold) {
  // 【step1】 
  // 左侧图片下载，转为blob。此操作只需要在每次点击row时执行。
  // 创建worker对象
  const workerFun = await spawn(new Worker("@/worker/ThreadWorkerMulti"))
  const blob = await workerFun.getImageBlob(leftImg)
  await Thread.terminate(workerFun);
  const start = performance.now();
  const myTasks = []
  // 【step2】
  // 右侧图片相似度对比与计算
  const pool = Pool(() => spawn(new Worker("@/worker/ThreadWorkerMulti")), {
    concurrency: 4,
    name: "task-match_img"
  })
  for (let i = 0; i < searchImgList.length; i++) {
    let matchData =searchImgList[i];
    const task = pool.queue(worker => worker.doImgDiff(blob, matchData.imgUrl, matchData.listId, threshold))
      .then()
      .catch((error) => {
        console.log("error ==", error)
      })
      .finally();
    myTasks.push(task);
  }
  const result = await Promise.all(myTasks)
    .finally(() => {
      // 关闭线程池
      pool.completed();
      pool.terminate();
    })

  const time  = performance.now() - start;
  console.log('花费时间:'+time)
  console.log("result ==", result)
  return result;
}

worker线程

// ThreadWorkerMulti.js
/* eslint-disable */
import {expose} from "threads"
import pixelmatch from 'pixelmatch'

/**
 *
 * @param tmpBlob 源图片blob
 * @param img2Url 要比对的图片url地址
 * @param listId  唯一标识id
 * @param threshold 系数
 */
async function doImgDiff(tmpBlob, img2Url, listId, threshold) {
    // 源图片，即左侧图片
    // const tmpBlob = await getImageBlob(img1Url)
    // 目标图片，即右侧分词查询list中的图片
    const tmpBlob2 = await getImageBlob(img2Url)
    const imageBitmap = await self.createImageBitmap(tmpBlob);
    const imageBitmap2 = await self.createImageBitmap(tmpBlob2);

    let canvas = new OffscreenCanvas(imageBitmap.width, imageBitmap.height);
        // willReadFrequently == > Canvas2D: Multiple readback operations using getImageData are faster with the willReadFrequently attribute set to true warnings
    let ctx = canvas.getContext('2d',{ willReadFrequently: true});
    const diff = ctx.createImageData(imageBitmap.width, imageBitmap.height)
    const imageData = getImageData(imageBitmap, ctx, canvas);
    const imageData2 = getImageData(imageBitmap2, ctx, canvas);

    const score = getImgScore(imageData, imageData2, diff, threshold, listId)
    return score;
}


/**
 * 获取图片blob二进制流
 * @param url
 * @returns {Promise<unknown>}
 */
function getImageBlob(url) {
    return new Promise((resolve, reject) => {
        const xhr = new XMLHttpRequest();
        url = "https://mp-statistics-dev.hanpda.com/tools/mstats/api/product/images?imageUrl=" + url
        xhr.open("get", url, true);
        xhr.responseType = "blob";
        xhr.onload = function () {
            if (this.status == 200) {
                resolve(this.response);
            } else {
                reject(new Error('error-'));
            }
        };
        xhr.send();
        xhr.onerror = () => {
            reject(new Error('error-'));
        }
    })
}

/**
 * 计算图片相似度
 * @param imageData1
 * @param imageData2
 * @param diff
 * @returns {number|number}
 */
function getImgScore(imageData1, imageData2, diff, _threshold, _listId) {
    const score = pixelmatch(imageData1.data, imageData2.data, diff.data, imageData1.width, imageData1.height, {threshold: _threshold})
    const percent = ~~(score / (imageData1.width * imageData1.height) * 1000) / 10
    const result = {
        listId: _listId,
        percent: percent
    }
    console.log("result ===", result)
    return result;
}

function getImageData(imageBitmap, ctx, canvas) {
    ctx.drawImage(imageBitmap, 0, 0, canvas.width, canvas.height)
    const imageData = ctx.getImageData(0, 0, canvas.width, canvas.height)
    return imageData;
}

// 要导出的函数
const exposeFun={
    doImgDiff, getImageBlob
}

expose(exposeFun)

Ⅱ. resemble + thread.js

1. 依赖安装与配置

shell

# browser浏览器运行不需要安装 node-canvas
npm install resemblejs
npm install --no-optional
# --------------------------------

# 在vue 文件中引入resemble.js [源码修改版]
# 在/public文件夹中引入 resemble.js
# 在 index.html中引入resemble.js
<script src="<%= BASE_URL %>/resemble.js"></script>

Resemble.js -- 修改版

javascript

// Resemble.js 源码修改版 -- canvas 修改为 offscreenCanvas,以ImageBitmap的方式处理图片
/*
James Cryer / Huddle
URL: https://github.com/Huddle/Resemble.js
*/

var naiveFallback = function () {
    // ISC (c) 2011-2019 https://github.com/medikoo/es5-ext/blob/master/global.js
    if (typeof self === "object" && self) {
        return self;
    }
    if (typeof window === "object" && window) {
        return window;
    }
    throw new Error("Unable to resolve global `this`");
};

var getGlobalThis = function () {
    // ISC (c) 2011-2019 https://github.com/medikoo/es5-ext/blob/master/global.js
    // Fallback to standard globalThis if available
    if (typeof globalThis === "object" && globalThis) {
        return globalThis;
    }

    try {
        Object.defineProperty(Object.prototype, "__global__", {
            get: function () {
                return this;
            },
            configurable: true
        });
    } catch (error) {
        return naiveFallback();
    }
    try {
        // eslint-disable-next-line no-undef
        if (!__global__) {
            return naiveFallback();
        }
        return __global__; // eslint-disable-line no-undef
    } finally {
        delete Object.prototype.__global__;
    }
};

var isNode = function () {
    const globalPolyfill = getGlobalThis();
    return typeof globalPolyfill.process !== "undefined" && globalPolyfill.process.versions && globalPolyfill.process.versions.node;
};

(function (root, factory) {
    "use strict";
    if (typeof define === "function" && define.amd) {
        define([], factory);
    } else if (typeof module === "object" && module.exports) {
        module.exports = factory();
    } else {
        root.resemble = factory();
    }
})(this /* eslint-disable-line no-invalid-this*/, function () {
    "use strict";

    var Img;
    var Canvas;
    var loadNodeCanvasImage;

    function createCanvas(width, height) {
      /*  if (isNode()) {
            return Canvas.createCanvas(width, height);
        }*/
        // console.log(" 创建离屏canvas")
        var cnvs = new OffscreenCanvas(width,height)

        return cnvs;
    }

    var oldGlobalSettings = {};
    var globalOutputSettings = oldGlobalSettings;

    var resemble = function (fileData) {
        var pixelTransparency = 1;

        var errorPixelColor = {
            // Color for Error Pixels. Between 0 and 255.
            red: 255,
            green: 0,
            blue: 255,
            alpha: 255
        };

        var targetPix = { r: 0, g: 0, b: 0, a: 0 }; // isAntialiased

        var errorPixelTransform = {
            flat: function (px, offset) {
                px[offset] = errorPixelColor.red;
                px[offset + 1] = errorPixelColor.green;
                px[offset + 2] = errorPixelColor.blue;
                px[offset + 3] = errorPixelColor.alpha;
            },
            movement: function (px, offset, d1, d2) {
                px[offset] = (d2.r * (errorPixelColor.red / 255) + errorPixelColor.red) / 2;
                px[offset + 1] = (d2.g * (errorPixelColor.green / 255) + errorPixelColor.green) / 2;
                px[offset + 2] = (d2.b * (errorPixelColor.blue / 255) + errorPixelColor.blue) / 2;
                px[offset + 3] = d2.a;
            },
            flatDifferenceIntensity: function (px, offset, d1, d2) {
                px[offset] = errorPixelColor.red;
                px[offset + 1] = errorPixelColor.green;
                px[offset + 2] = errorPixelColor.blue;
                px[offset + 3] = colorsDistance(d1, d2);
            },
            movementDifferenceIntensity: function (px, offset, d1, d2) {
                var ratio = (colorsDistance(d1, d2) / 255) * 0.8;

                px[offset] = (1 - ratio) * (d2.r * (errorPixelColor.red / 255)) + ratio * errorPixelColor.red;
                px[offset + 1] = (1 - ratio) * (d2.g * (errorPixelColor.green / 255)) + ratio * errorPixelColor.green;
                px[offset + 2] = (1 - ratio) * (d2.b * (errorPixelColor.blue / 255)) + ratio * errorPixelColor.blue;
                px[offset + 3] = d2.a;
            },
            diffOnly: function (px, offset, d1, d2) {
                px[offset] = d2.r;
                px[offset + 1] = d2.g;
                px[offset + 2] = d2.b;
                px[offset + 3] = d2.a;
            }
        };

        var errorPixel = errorPixelTransform.flat;
        var errorType;
        var boundingBoxes;
        var ignoredBoxes;
        var ignoreAreasColoredWith;
        var largeImageThreshold = 1200;
        var useCrossOrigin = true;
        var data = {};
        var images = [];
        var updateCallbackArray = [];

        var tolerance = {
            // between 0 and 255
            red: 16,
            green: 16,
            blue: 16,
            alpha: 16,
            minBrightness: 16,
            maxBrightness: 240
        };

        var ignoreAntialiasing = false;
        var ignoreColors = false;
        var scaleToSameSize = false;
        var compareOnly = false;
        var returnEarlyThreshold;

        function colorsDistance(c1, c2) {
            return (Math.abs(c1.r - c2.r) + Math.abs(c1.g - c2.g) + Math.abs(c1.b - c2.b)) / 3;
        }

        function withinBoundingBox(x, y, width, height, box) {
            return x > (box.left || 0) && x < (box.right || width) && y > (box.top || 0) && y < (box.bottom || height);
        }

        function withinComparedArea(x, y, width, height, pixel2) {
            var isIncluded = true;
            var i;
            var boundingBox;
            var ignoredBox;
            var selected;
            var ignored;

            if (boundingBoxes instanceof Array) {
                selected = false;
                for (i = 0; i < boundingBoxes.length; i++) {
                    boundingBox = boundingBoxes[i];
                    if (withinBoundingBox(x, y, width, height, boundingBox)) {
                        selected = true;
                        break;
                    }
                }
            }
            if (ignoredBoxes instanceof Array) {
                ignored = true;
                for (i = 0; i < ignoredBoxes.length; i++) {
                    ignoredBox = ignoredBoxes[i];
                    if (withinBoundingBox(x, y, width, height, ignoredBox)) {
                        ignored = false;
                        break;
                    }
                }
            }

            if (ignoreAreasColoredWith) {
                return colorsDistance(pixel2, ignoreAreasColoredWith) !== 0;
            }

            if (selected === undefined && ignored === undefined) {
                return true;
            }
            if (selected === false && ignored === true) {
                return false;
            }
            if (selected === true || ignored === true) {
                isIncluded = true;
            }
            if (selected === false || ignored === false) {
                isIncluded = false;
            }
            return isIncluded;
        }

        function triggerDataUpdate() {
            var len = updateCallbackArray.length;
            var i;
            for (i = 0; i < len; i++) {
                if (typeof updateCallbackArray[i] === "function") {
                    updateCallbackArray[i](data);
                }
            }
        }

        function loop(w, h, callback) {
            var x;
            var y;

            for (x = 0; x < w; x++) {
                for (y = 0; y < h; y++) {
                    callback(x, y);
                }
            }
        }

        function parseImage(sourceImageData, width, height) {
            var pixelCount = 0;
            var redTotal = 0;
            var greenTotal = 0;
            var blueTotal = 0;
            var alphaTotal = 0;
            var brightnessTotal = 0;
            var whiteTotal = 0;
            var blackTotal = 0;

            loop(width, height, function (horizontalPos, verticalPos) {
                var offset = (verticalPos * width + horizontalPos) * 4;
                var red = sourceImageData[offset];
                var green = sourceImageData[offset + 1];
                var blue = sourceImageData[offset + 2];
                var alpha = sourceImageData[offset + 3];
                var brightness = getBrightness(red, green, blue);

                if (red === green && red === blue && alpha) {
                    if (red === 0) {
                        blackTotal++;
                    } else if (red === 255) {
                        whiteTotal++;
                    }
                }

                pixelCount++;

                redTotal += (red / 255) * 100;
                greenTotal += (green / 255) * 100;
                blueTotal += (blue / 255) * 100;
                alphaTotal += ((255 - alpha) / 255) * 100;
                brightnessTotal += (brightness / 255) * 100;
            });

            data.red = Math.floor(redTotal / pixelCount);
            data.green = Math.floor(greenTotal / pixelCount);
            data.blue = Math.floor(blueTotal / pixelCount);
            data.alpha = Math.floor(alphaTotal / pixelCount);
            data.brightness = Math.floor(brightnessTotal / pixelCount);
            data.white = Math.floor((whiteTotal / pixelCount) * 100);
            data.black = Math.floor((blackTotal / pixelCount) * 100);

            triggerDataUpdate();
        }

        function onLoadImage(hiddenImage, callback) {
            // don't assign to hiddenImage, see https://github.com/Huddle/Resemble.js/pull/87/commits/300d43352a2845aad289b254bfbdc7cd6a37e2d7
            var width = hiddenImage.width;
            var height = hiddenImage.height;
            if (scaleToSameSize && images.length === 1) {
                width = images[0].width;
                height = images[0].height;
            }

            var hiddenCanvas = createCanvas(width, height);
            var imageData;

            hiddenCanvas.getContext("2d").drawImage(hiddenImage, 0, 0, width, height);
            imageData = hiddenCanvas.getContext("2d").getImageData(0, 0, width, height);

            images.push(imageData);

            callback(imageData, width, height);
        }

        function loadImageData(fileDataForImage, callback) {
            onLoadImage(fileDataForImage, callback);
        }

        function isColorSimilar(a, b, color) {
            var absDiff = Math.abs(a - b);

            if (typeof a === "undefined") {
                return false;
            }
            if (typeof b === "undefined") {
                return false;
            }

            if (a === b) {
                return true;
            } else if (absDiff < tolerance[color]) {
                return true;
            }
            return false;
        }

        function isPixelBrightnessSimilar(d1, d2) {
            var alpha = isColorSimilar(d1.a, d2.a, "alpha");
            var brightness = isColorSimilar(d1.brightness, d2.brightness, "minBrightness");
            return brightness && alpha;
        }

        function getBrightness(r, g, b) {
            return 0.3 * r + 0.59 * g + 0.11 * b;
        }

        function isRGBSame(d1, d2) {
            var red = d1.r === d2.r;
            var green = d1.g === d2.g;
            var blue = d1.b === d2.b;
            return red && green && blue;
        }

        function isRGBSimilar(d1, d2) {
            var red = isColorSimilar(d1.r, d2.r, "red");
            var green = isColorSimilar(d1.g, d2.g, "green");
            var blue = isColorSimilar(d1.b, d2.b, "blue");
            var alpha = isColorSimilar(d1.a, d2.a, "alpha");

            return red && green && blue && alpha;
        }

        function isContrasting(d1, d2) {
            return Math.abs(d1.brightness - d2.brightness) > tolerance.maxBrightness;
        }

        function getHue(red, green, blue) {
            var r = red / 255;
            var g = green / 255;
            var b = blue / 255;
            var max = Math.max(r, g, b);
            var min = Math.min(r, g, b);
            var h;
            var d;

            if (max === min) {
                h = 0; // achromatic
            } else {
                d = max - min;
                switch (max) {
                    case r:
                        h = (g - b) / d + (g < b ? 6 : 0);
                        break;
                    case g:
                        h = (b - r) / d + 2;
                        break;
                    case b:
                        h = (r - g) / d + 4;
                        break;
                    default:
                        h /= 6;
                }
            }

            return h;
        }

        function isAntialiased(sourcePix, pix, cacheSet, verticalPos, horizontalPos, width) {
            var offset;
            var distance = 1;
            var i;
            var j;
            var hasHighContrastSibling = 0;
            var hasSiblingWithDifferentHue = 0;
            var hasEquivalentSibling = 0;

            addHueInfo(sourcePix);

            for (i = distance * -1; i <= distance; i++) {
                for (j = distance * -1; j <= distance; j++) {
                    if (i === 0 && j === 0) {
                        // ignore source pixel
                    } else {
                        offset = ((verticalPos + j) * width + (horizontalPos + i)) * 4;

                        if (!getPixelInfo(targetPix, pix, offset, cacheSet)) {
                            continue;
                        }

                        addBrightnessInfo(targetPix);
                        addHueInfo(targetPix);

                        if (isContrasting(sourcePix, targetPix)) {
                            hasHighContrastSibling++;
                        }

                        if (isRGBSame(sourcePix, targetPix)) {
                            hasEquivalentSibling++;
                        }

                        if (Math.abs(targetPix.h - sourcePix.h) > 0.3) {
                            hasSiblingWithDifferentHue++;
                        }

                        if (hasSiblingWithDifferentHue > 1 || hasHighContrastSibling > 1) {
                            return true;
                        }
                    }
                }
            }

            if (hasEquivalentSibling < 2) {
                return true;
            }

            return false;
        }

        function copyPixel(px, offset, pix) {
            if (errorType === "diffOnly") {
                return;
            }

            px[offset] = pix.r; // r
            px[offset + 1] = pix.g; // g
            px[offset + 2] = pix.b; // b
            px[offset + 3] = pix.a * pixelTransparency; // a
        }

        function copyGrayScalePixel(px, offset, pix) {
            if (errorType === "diffOnly") {
                return;
            }

            px[offset] = pix.brightness; // r
            px[offset + 1] = pix.brightness; // g
            px[offset + 2] = pix.brightness; // b
            px[offset + 3] = pix.a * pixelTransparency; // a
        }

        function getPixelInfo(dst, pix, offset) {
            if (pix.length > offset) {
                dst.r = pix[offset];
                dst.g = pix[offset + 1];
                dst.b = pix[offset + 2];
                dst.a = pix[offset + 3];

                return true;
            }

            return false;
        }

        function addBrightnessInfo(pix) {
            pix.brightness = getBrightness(pix.r, pix.g, pix.b); // 'corrected' lightness
        }

        function addHueInfo(pix) {
            pix.h = getHue(pix.r, pix.g, pix.b);
        }

        function analyseImages(img1, img2, width, height) {
            var data1 = img1.data;
            var data2 = img2.data;
            var hiddenCanvas;
            var context;
            var imgd;
            var pix;

            if (!compareOnly) {
                hiddenCanvas = createCanvas(width, height);

                context = hiddenCanvas.getContext("2d");
                imgd = context.createImageData(width, height);
                pix = imgd.data;
            }

            var mismatchCount = 0;
            var diffBounds = {
                top: height,
                left: width,
                bottom: 0,
                right: 0
            };
            var updateBounds = function (x, y) {
                diffBounds.left = Math.min(x, diffBounds.left);
                diffBounds.right = Math.max(x, diffBounds.right);
                diffBounds.top = Math.min(y, diffBounds.top);
                diffBounds.bottom = Math.max(y, diffBounds.bottom);
            };

            var time = Date.now();

            var skip;

            if (!!largeImageThreshold && ignoreAntialiasing && (width > largeImageThreshold || height > largeImageThreshold)) {
                skip = 6;
            }

            var pixel1 = { r: 0, g: 0, b: 0, a: 0 };
            var pixel2 = { r: 0, g: 0, b: 0, a: 0 };

            var skipTheRest = false;

            loop(width, height, function (horizontalPos, verticalPos) {
                if (skipTheRest) {
                    return;
                }

                if (skip) {
                    // only skip if the image isn't small
                    if (verticalPos % skip === 0 || horizontalPos % skip === 0) {
                        return;
                    }
                }

                var offset = (verticalPos * width + horizontalPos) * 4;
                if (!getPixelInfo(pixel1, data1, offset, 1) || !getPixelInfo(pixel2, data2, offset, 2)) {
                    return;
                }

                var isWithinComparedArea = withinComparedArea(horizontalPos, verticalPos, width, height, pixel2);

                if (ignoreColors) {
                    addBrightnessInfo(pixel1);
                    addBrightnessInfo(pixel2);

                    if (isPixelBrightnessSimilar(pixel1, pixel2) || !isWithinComparedArea) {
                        if (!compareOnly) {
                            copyGrayScalePixel(pix, offset, pixel2);
                        }
                    } else {
                        if (!compareOnly) {
                            errorPixel(pix, offset, pixel1, pixel2);
                        }

                        mismatchCount++;
                        updateBounds(horizontalPos, verticalPos);
                    }
                    return;
                }

                if (isRGBSimilar(pixel1, pixel2) || !isWithinComparedArea) {
                    if (!compareOnly) {
                        copyPixel(pix, offset, pixel1);
                    }
                } else if (
                    ignoreAntialiasing &&
                    (addBrightnessInfo(pixel1), // jit pixel info augmentation looks a little weird, sorry.
                    addBrightnessInfo(pixel2),
                    isAntialiased(pixel1, data1, 1, verticalPos, horizontalPos, width) || isAntialiased(pixel2, data2, 2, verticalPos, horizontalPos, width))
                ) {
                    if (isPixelBrightnessSimilar(pixel1, pixel2) || !isWithinComparedArea) {
                        if (!compareOnly) {
                            copyGrayScalePixel(pix, offset, pixel2);
                        }
                    } else {
                        if (!compareOnly) {
                            errorPixel(pix, offset, pixel1, pixel2);
                        }

                        mismatchCount++;
                        updateBounds(horizontalPos, verticalPos);
                    }
                } else {
                    if (!compareOnly) {
                        errorPixel(pix, offset, pixel1, pixel2);
                    }

                    mismatchCount++;
                    updateBounds(horizontalPos, verticalPos);
                }

                if (compareOnly) {
                    var currentMisMatchPercent = (mismatchCount / (height * width)) * 100;

                    if (currentMisMatchPercent > returnEarlyThreshold) {
                        skipTheRest = true;
                    }
                }
            });

            data.rawMisMatchPercentage = (mismatchCount / (height * width)) * 100;
            data.misMatchPercentage = data.rawMisMatchPercentage.toFixed(2);
            data.diffBounds = diffBounds;
            data.analysisTime = Date.now() - time;

            data.getImageDataUrl = function (text) {
                if (compareOnly) {
                    throw Error("No diff image available - ran in compareOnly mode");
                }

                var barHeight = 0;

                if (text) {
                    barHeight = addLabel(text, context, hiddenCanvas);
                }

                context.putImageData(imgd, 0, barHeight);

                return hiddenCanvas.toDataURL("image/png");
            };

            if (!compareOnly && hiddenCanvas.toBuffer) {
                data.getBuffer = function (includeOriginal) {
                    if (includeOriginal) {
                        var imageWidth = hiddenCanvas.width + 2;
                        hiddenCanvas.width = imageWidth * 3;
                        context.putImageData(img1, 0, 0);
                        context.putImageData(img2, imageWidth, 0);
                        context.putImageData(imgd, imageWidth * 2, 0);
                    } else {
                        context.putImageData(imgd, 0, 0);
                    }
                    return hiddenCanvas.toBuffer();
                };
            }
        }

        function addLabel(text, context, hiddenCanvas) {
            var textPadding = 2;

            context.font = "12px sans-serif";

            var textWidth = context.measureText(text).width + textPadding * 2;
            var barHeight = 22;

            if (textWidth > hiddenCanvas.width) {
                hiddenCanvas.width = textWidth;
            }

            hiddenCanvas.height += barHeight;

            context.fillStyle = "#666";
            context.fillRect(0, 0, hiddenCanvas.width, barHeight - 4);
            context.fillStyle = "#fff";
            context.fillRect(0, barHeight - 4, hiddenCanvas.width, 4);

            context.fillStyle = "#fff";
            context.textBaseline = "top";
            context.font = "12px sans-serif";
            context.fillText(text, textPadding, 1);

            return barHeight;
        }

        function normalise(img, w, h) {
            var c;
            var context;

            if (img.height < h || img.width < w) {
                c = createCanvas(w, h);
                context = c.getContext("2d");
                context.putImageData(img, 0, 0);
                return context.getImageData(0, 0, w, h);
            }

            return img;
        }

        function outputSettings(options) {
            var key;

            if (options.errorColor) {
                for (key in options.errorColor) {
                    if (options.errorColor.hasOwnProperty(key)) {
                        errorPixelColor[key] = options.errorColor[key] === void 0 ? errorPixelColor[key] : options.errorColor[key];
                    }
                }
            }

            if (options.errorType && errorPixelTransform[options.errorType]) {
                errorPixel = errorPixelTransform[options.errorType];
                errorType = options.errorType;
            }

            if (options.errorPixel && typeof options.errorPixel === "function") {
                errorPixel = options.errorPixel;
            }

            pixelTransparency = isNaN(Number(options.transparency)) ? pixelTransparency : options.transparency;

            if (options.largeImageThreshold !== undefined) {
                largeImageThreshold = options.largeImageThreshold;
            }

            if (options.useCrossOrigin !== undefined) {
                useCrossOrigin = options.useCrossOrigin;
            }

            if (options.boundingBox !== undefined) {
                boundingBoxes = [options.boundingBox];
            }

            if (options.ignoredBox !== undefined) {
                ignoredBoxes = [options.ignoredBox];
            }

            if (options.boundingBoxes !== undefined) {
                boundingBoxes = options.boundingBoxes;
            }

            if (options.ignoredBoxes !== undefined) {
                ignoredBoxes = options.ignoredBoxes;
            }

            if (options.ignoreAreasColoredWith !== undefined) {
                ignoreAreasColoredWith = options.ignoreAreasColoredWith;
            }
        }

        function compare(one, two) {
            if (globalOutputSettings !== oldGlobalSettings) {
                outputSettings(globalOutputSettings);
            }

            function onceWeHaveBoth() {
                var width;
                var height;
                if (images.length === 2) {
                    if (images[0].error || images[1].error) {
                        data = {};
                        data.error = images[0].error ? images[0].error : images[1].error;
                        triggerDataUpdate();
                        return;
                    }
                    width = images[0].width > images[1].width ? images[0].width : images[1].width;
                    height = images[0].height > images[1].height ? images[0].height : images[1].height;

                    if (images[0].width === images[1].width && images[0].height === images[1].height) {
                        data.isSameDimensions = true;
                    } else {
                        data.isSameDimensions = false;
                    }

                    data.dimensionDifference = {
                        width: images[0].width - images[1].width,
                        height: images[0].height - images[1].height
                    };

                    analyseImages(normalise(images[0], width, height), normalise(images[1], width, height), width, height);

                    triggerDataUpdate();
                }
            }

            images = [];
            loadImageData(one, onceWeHaveBoth);
            loadImageData(two, onceWeHaveBoth);
        }

        function getCompareApi(param) {
            var secondFileData;
            var hasMethod = typeof param === "function";

            if (!hasMethod) {
                // assume it's file data
                secondFileData = param;
            }

            var self = {
                setReturnEarlyThreshold: function (threshold) {
                    if (threshold) {
                        compareOnly = true;
                        returnEarlyThreshold = threshold;
                    }
                    return self;
                },
                scaleToSameSize: function () {
                    scaleToSameSize = true;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                useOriginalSize: function () {
                    scaleToSameSize = false;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                ignoreNothing: function () {
                    tolerance.red = 0;
                    tolerance.green = 0;
                    tolerance.blue = 0;
                    tolerance.alpha = 0;
                    tolerance.minBrightness = 0;
                    tolerance.maxBrightness = 255;

                    ignoreAntialiasing = false;
                    ignoreColors = false;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                ignoreLess: function () {
                    tolerance.red = 16;
                    tolerance.green = 16;
                    tolerance.blue = 16;
                    tolerance.alpha = 16;
                    tolerance.minBrightness = 16;
                    tolerance.maxBrightness = 240;

                    ignoreAntialiasing = false;
                    ignoreColors = false;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                ignoreAntialiasing: function () {
                    tolerance.red = 32;
                    tolerance.green = 32;
                    tolerance.blue = 32;
                    tolerance.alpha = 32;
                    tolerance.minBrightness = 64;
                    tolerance.maxBrightness = 96;

                    ignoreAntialiasing = true;
                    ignoreColors = false;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                ignoreColors: function () {
                    tolerance.alpha = 16;
                    tolerance.minBrightness = 16;
                    tolerance.maxBrightness = 240;

                    ignoreAntialiasing = false;
                    ignoreColors = true;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                ignoreAlpha: function () {
                    tolerance.red = 16;
                    tolerance.green = 16;
                    tolerance.blue = 16;
                    tolerance.alpha = 255;
                    tolerance.minBrightness = 16;
                    tolerance.maxBrightness = 240;

                    ignoreAntialiasing = false;
                    ignoreColors = false;

                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                repaint: function () {
                    if (hasMethod) {
                        param();
                    }
                    return self;
                },
                outputSettings: function (options) {
                    outputSettings(options);
                    return self;
                },
                onComplete: function (callback) {
                    updateCallbackArray.push(callback);

                    var wrapper = function () {
                        compare(fileData, secondFileData);
                    };

                    wrapper();

                    return getCompareApi(wrapper);
                },
                setupCustomTolerance: function (customSettings) {
                    for (var property in tolerance) {
                        if (!customSettings.hasOwnProperty(property)) {
                            continue;
                        }

                        tolerance[property] = customSettings[property];
                    }
                }
            };

            return self;
        }

        var rootSelf = {
            onComplete: function (callback) {
                updateCallbackArray.push(callback);
                loadImageData(fileData, function (imageData, width, height) {
                    parseImage(imageData.data, width, height);
                });
            },
            compareTo: function (secondFileData) {
                return getCompareApi(secondFileData);
            },
            outputSettings: function (options) {
                outputSettings(options);
                return rootSelf;
            }
        };

        return rootSelf;
    };

    function setGlobalOutputSettings(settings) {
        globalOutputSettings = settings;
        return resemble;
    }

    function applyIgnore(api, ignore, customTolerance) {
        switch (ignore) {
            case "nothing":
                api.ignoreNothing();
                break;
            case "less":
                api.ignoreLess();
                break;
            case "antialiasing":
                api.ignoreAntialiasing();
                break;
            case "colors":
                api.ignoreColors();
                break;
            case "alpha":
                api.ignoreAlpha();
                break;
            default:
                throw new Error("Invalid ignore: " + ignore);
        }

        api.setupCustomTolerance(customTolerance);
    }

    resemble.compare = function (image1, image2, options, cb) {
        var callback;
        var opt;

        if (typeof options === "function") {
            callback = options;
            opt = {};
        } else {
            callback = cb;
            opt = options || {};
        }

        var res = resemble(image1);
        var compare;

        if (opt.output) {
            res.outputSettings(opt.output);
        }

        compare = res.compareTo(image2);

        if (opt.returnEarlyThreshold) {
            compare.setReturnEarlyThreshold(opt.returnEarlyThreshold);
        }

        if (opt.scaleToSameSize) {
            compare.scaleToSameSize();
        }

        var toleranceSettings = opt.tolerance || {};
        if (typeof opt.ignore === "string") {
            applyIgnore(compare, opt.ignore, toleranceSettings);
        } else if (opt.ignore && opt.ignore.forEach) {
            opt.ignore.forEach(function (v) {
                applyIgnore(compare, v, toleranceSettings);
            });
        }

        compare.onComplete(function (data) {
            if (data.error) {
                callback(data.error);
            } else {
                callback(null, data);
            }
        });
    };

    resemble.outputSettings = setGlobalOutputSettings;
    return resemble;
});

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007

2. 业务逻辑

main线程

javascript

// MatchingSearch.vue
// 触发方法
const leftImg =await this.current.mainImage;
const searchDataList = [];
results.forEach(item => {
  const tmpData = {
    custom: {
      leftImg: leftImg,
      rightImg: item.ecImageUrl
    },
    imgUrl: item.ecImageUrl
  }
  searchDataList.push(tmpData)
})
this.imgDiffTaskResembleTask(leftImg, searchDataList)


/**
 * 实际方法
 **/
async imgDiffTaskResembleTask(leftImg, searchImgList){
  console.log("searchImgList ==", searchImgList.length)
  const workerFun = await spawn(new Worker("@/worker/RemblThreadWorker"))
  const img1Blob = await workerFun.getImageBlob(leftImg)
  await Thread.terminate(workerFun);
  const start = performance.now();
  const myTasks = []
  const pool = Pool(() => spawn(new Worker("@/worker/RemblThreadWorker")), {
    concurrency: 4,
    name: "task-match_img_rembl"
  })
  let len = searchImgList.length;
  for (let i = 0; i < len; i++) {
    let matchData = searchImgList[i];
   const task = pool.queue(worker => worker.getRemblMisMatchPercentage(img1Blob, matchData.imgUrl, matchData.custom))
      .then()
      .catch((error) => {
        console.log("error ==", error)
      })
      .finally();
    myTasks.push(task);
  }
  const result = await Promise.all(myTasks)
    .finally(() => {
      // 关闭线程池
      pool.completed();
      pool.terminate();
    })

  const time = performance.now() - start;
  console.log('花费时间:' + time)
  console.log("result ==", result)
  return result;
}

worker线程

javascript

// RemblThreadWorker.js
/* eslint-disable */
import {expose} from "threads"
// 导入 /public下的resemble.js文件
importScripts("../resemble.js")


/**
 * 获取图片分析结果
 * @param analysisPng
 * @returns {Promise<unknown>}
 */
function getRemblAnalysisInfo(analysisPng) {
    return new Promise(resolve => {
        resemble(analysisPng).onComplete(function (data) {
            resolve(data)
        })
    })
}

/**
 * 查询图片比对相似度
 * @param blob1
 * @param img2Url
 * @param custom
 * @returns {Promise<unknown>}
 */
async function getRemblMisMatchPercentage(blob1, img2Url, custom) {
    const blob2 = await getImageBlob(img2Url);
    const imageBitmap = await self.createImageBitmap(blob1);
    const image2Bitmap = await self.createImageBitmap(blob2);
    const options = await getRemblOptions(imageBitmap, image2Bitmap);
    return new Promise(resolve => {
        resemble(imageBitmap)
            .compareTo(image2Bitmap)
           /* .ignoreNothing()*/
            // 使用相同尺寸
            .scaleToSameSize()
            // 忽略抗锯齿 （位置不同，便于查找）
            .ignoreAntialiasing()
            // 个性化参数配置
            .outputSettings(options)
            .onComplete((data) => {
                //console.log("不同点 :" + data.misMatchPercentage + "%")
                const result = {
                    percentage: (100 - data.misMatchPercentage).toFixed(2),
                    custom: custom
                }
                resolve(result);
            });
    })
}

/**
 * 设置rembl option配置
 * @param blob1
 * @param blob2
 * @returns {Promise<{}>}
 */
async function getRemblOptions(blob1, blob2) {
    const options = {outputDiff: false};
/*    // 忽略白色
    const ignoreAreasColoredWith = {
        r: 255,
        g: 255,
        b: 255,
        a: 0
    }

    // 分析第二张图片，他的white像素占比情况，如果占比>55%,那么比较图片时不可以忽略while底色。
    const analysisInfo = await getRemblAnalysisInfo(blob2)
    const white = analysisInfo.white;
    const tmpArr = [];
    tmpArr.push(analysisInfo.alpha)
    tmpArr.push(analysisInfo.black)
    tmpArr.push(analysisInfo.blue)
    tmpArr.push(analysisInfo.brightness)
    tmpArr.push(analysisInfo.green)
    tmpArr.push(analysisInfo.red)
    const max = Math.max(...tmpArr);
    // 排除底色 white(255,255,255) 导致的 图片特别相似的问题
    if (white / max < 0.55) {
        options.ignoreAreasColoredWith = ignoreAreasColoredWith;
    }
    options.largeImageThreshold = 0
    options.outputDiff = false*/
    return options;
}

/**
 * 背景是否是白色
 */
async function isBackgroundWhite(img) {
    const analysisInfo = await getRemblAnalysisInfo(blob2)
    const white = analysisInfo.white;
    const tmpArr = [];
    tmpArr.push(analysisInfo.alpha)
    tmpArr.push(analysisInfo.black)
    tmpArr.push(analysisInfo.blue)
    tmpArr.push(analysisInfo.brightness)
    tmpArr.push(analysisInfo.green)
    tmpArr.push(analysisInfo.red)
    const max = Math.max(...tmpArr);
    // 排除底色 white(255,255,255) 导致的 图片特别相似的问题
    return (white/max).toFixed(2);
}


/**
 * 获取图片blob二进制流
 * @param url
 * @returns {Promise<unknown>}
 */
function getImageBlob(url) {
    return new Promise((resolve, reject) => {
        const xhr = new XMLHttpRequest();
        url = "https://mp-statistics-dev.hanpda.com/tools/mstats/api/product/images?imageUrl=" + url
        xhr.open("get", url, true);
        xhr.responseType = "blob";
        xhr.onload = function () {
            if (this.status == 200) {
                resolve(this.response);
            } else {
                reject(new Error('error-'));
            }
        };
        xhr.send();
        xhr.onerror = (error) => {
            reject(new Error('error-'));
        }
    })
}

// 要导出的函数
const exposeFun = {getImageBlob, getRemblMisMatchPercentage}

expose(exposeFun)

性能测试

多核与性能的关系

使用nginx做代理，转发图片请求，解决跨域问题。
pixelmatch 参数配置 theshold = 0.1;
resemble 使用默认配置，只开启resize；
每种场景耗时均获取10次请求的平均值计算而来

Times	Source Size	cpu cores	Resemble (ms)	Pixelmatch (ms)
1904	119M	12	24307	24605
1904	119M	10	24083	43015
1904	119M	8	24250	43209
1904	119M	6	28744	48946
1904	119M	4	31550	60546

结论

综合来看 resemble 的性能要优于pixelmatch。尤其是多线程场景下。从上述数据结果分析，我们不难发现，性能限制与浏览器tcp同一时间内最大是6个的限制基本一致。推荐配置6核，性能配置8核。

二、java图片相似度计算

使用JImageHash依赖库来实现图片的相似度比对，然后计算hamming距离来标识相似度
实际测试多线程下性能较差，不考虑。

三、nodejs 图片相似度计算

使用express框架作为web服务
pool：piscina + shuffled-priority-queue
算法库：imghash + hamming

依赖库

shell

# axios请求
npm install axios
# hash计算核心库
npm install blockhash-core
# 计算两个hash之间的hamming距离
npm install hamming
# 根据图片buffer生成hash
npm install imghash
# 线程池管理
npm install piscina
# 队列，与piscina配合使用

项目结构目录

shell

--public
--routes
  |--image-diff.js
--services
  |--ImgPiscinaService.js
--workers
  |--worker.js

实现

1. app配置router

javascript

...
// 引入image-diff router
var imgRouter = require('./routes/image-diff')
...
// 使用router路由
app.use('/img', imgRouter);

2. routes：image-diff.js

javascript

var express = require('express');
var router = express.Router();
const imgPiscinaService = require('../services/ImgPiscinaService')

router.get('/hash', async function (req, res, next) {
    const bb = await imgPiscinaService.getImgHash();
    res.send(bb);
});

// getHammingScore
router.get('/hamming', async function (req, res, next) {
    res.send(imgPiscinaService.getHammingScore());
});

module.exports = router;

3. services：ImgPiscinaService.js

javascript

const Piscina = require('piscina')
const path = require('path');
const spq = require('shuffled-priority-queue')
const fs = require("fs");
const kItem = Symbol('item');
const configFile = path.resolve(__dirname, '../mock.json')
const leftMockFile = path.resolve(__dirname, '../mock-left.json')

const hammingLeftFile = path.resolve(__dirname, '../mock-hamming-left.json')
const hammingRightFile = path.resolve(__dirname, '../mock-hamming-right.json')
const hamming = require('hamming')


class PriorityTaskQueue {
    queue = spq();

    get size() {
        return this.queue.length;
    }

    push(value) {
        const queueOptions = value[Piscina.queueOptionsSymbol];
        const priority = queueOptions ? (queueOptions.priority || 0) : 0;
        value[kItem] = this.queue.add({priority, value});
    }

    remove(value) {
        this.queue.remove(value[kItem]);
    }

    shift() {
        return this.queue.shift().value;
    }
}

/**
 * 使用Piscina 创建线程池pools
 * @type {Piscina}
 */
const pool = new Piscina({
    // worker文件path
    filename: path.resolve(__dirname, "../workers/worker.js"),
    // 队列，这里使用shuffled-priority-queue创建
    taskQueue: new PriorityTaskQueue(),
    // 单个线程的超时时间，1500ms。
    idleTimeout: 1500
})

/*const hammingPool = new Piscina({
    filename: path.resolve(__dirname, "../workers/hamming-worker.js"),
    taskQueue: new PriorityTaskQueue(),
    idleTimeout: 1500
})*/


const ImgService = {
    // 生成img hash
    /**
     *
     * @param leftList => { url:"http://10.15.1.175:80/img4/99.jpg" }
     * @param rightList => { url:"http://10.15.1.175:80/img4/99.jpg"}
     * @returns {Promise<{left: unknown[], right: unknown[]}>}
     */
    async getImgHash(leftList, rightList) {
        // mock data
        const leftData = fs.readFileSync(leftMockFile, 'UTF-8').toString();
        let leftJson = JSON.parse(leftData);
        leftList = leftJson;
        const data = fs.readFileSync(configFile, 'UTF-8').toString();
        let json = JSON.parse(data);
        rightList = json;

        // 计算图片hash值
        console.time("times")
        // 左侧图片的hash计算
        const leftArr = await this.doTask(leftList);
        // 右侧图片的hash计算
        const rightArr = await this.doTask(rightList);
        const result = {
            left: leftArr,
            right: rightArr
        }
        console.timeEnd("times")
        return result
    },
    /**
     * 计算hamming距离，找到左侧对比相似的图片
     * @param leftArr
     * @param rightArr
     * @returns {*[]}
     */
    getHammingScore(leftArr, rightArr) {
        // mock data
        const leftData = fs.readFileSync(hammingLeftFile, 'UTF-8').toString();
        let leftJson = JSON.parse(leftData);
        leftArr = leftJson;
        const rightData = fs.readFileSync(hammingRightFile, 'UTF-8').toString();
        let rightJson = JSON.parse(rightData);
        rightArr = rightJson;

        const result = [];
        const hashData = {
            leftObj: {},
            rightList: []
        }
        // 左侧列表，匹配右侧找到最相似的图像
        for (let i = 0; i < leftArr.length; i++) {
            const hash1 = leftArr[i].imageHashCode
            const rightList = [];
            // 编译右侧列表，计算hamming 距离
            for (let j = 0; j < rightArr.length; j++) {
                const hash2 = rightArr[j].imageHashCode;
                const score = hamming(hash1, hash2);
                if (score != null && score <= 5) {
                    const data = rightArr[j];
                    data.score = score;
                    rightList.push(data);
                }
            }
            if (rightList.length > 0) {
                hashData.leftObj = leftArr[i];
                hashData.rightList = rightList
                result.push(hashData);
            }
        }
        return result;
    },
    /**
     * 使用线程池添加任务，获取线程池任务执行结果
     * @param urlList
     * @returns {Promise<unknown[]>}
     */
    async doTask(urlList) {
        const rightTasks = [];
        for (let i = 0; i < urlList.length; i++) {
            let url = urlList[i].imageUrl;
            console.log("i" + i + "; url =" + url)
            const task = pool.run({imgUrl: url, custom: urlList[i]}, {name: 'getImageHash'})
            rightTasks.push(task);
        }
        const rightArr = await Promise.all(rightTasks)
        return rightArr;
    }
}

module.exports = ImgService

4. workers: worker.js

javascript

const imageHash = require('imghash')
const axios = require('axios')

const exportFun = {
    getImageHash({imgUrl, custom}) {
        return new Promise(resolve => {
            /**
             * 使用axios请求远程图片地址并获取buffer
             *
             * responseType: arraybuffer
             * timeout 1000: 最大等待时间为1000ms
             */
            axios({method: 'get', url: imgUrl, responseType: 'arraybuffer', timeout: 1000})
                .then(async (res) => {
                    /**
                     * 配置bits未16位。默认是8位，生成的hash过短，导致hamming计算时计算出的图片差异较大。
                     * @type {string | string}
                     */
                    const hash = await imageHash.hash(res.data, 16)
                    console.log("hash ==", hash)
                    custom.imageHashCode = hash
                    resolve(custom);
                }).catch(e => {
                console.log("error ==", e)
                custom.imageHashCode = ""
                resolve(custom);
            })
        })
    }
}


module.exports = exportFun;

性能

11423张图片，总图片尺寸119M。压缩后11M

nodejs 开启的线程依赖于cpu线程数（12核/线程）。

size	disk unzip (s)	disk zip (s)	url unzip (s)	url zip (s)
11423	28.630	2.330	42.127	6.152

四、golang图片相似度计算

golang的性能高、占用小。以线上服务为例，8协程 5600请求，3s内处理完毕。硬件：2核2g

核心代码

ImgService.go

package img_service

import (
	"bytes"
	"crypto/tls"
	"errors"
	"fmt"
	"github.com/corona10/goimagehash"
	"github.com/disintegration/imageorient"
	"github.com/gin-gonic/gin"
	"github.com/panjf2000/ants/v2"
	"github.com/valyala/fasthttp"
	"go-img/pkg/app"
	"go-img/pkg/e"
	"image"
	_ "image/gif"
	_ "image/jpeg"
	_ "image/png"
	"log"
	"net/http"
	"strconv"
	"sync"
	"time"
)

const (
	routineCountTotal = 8
)

var client *http.Client
var FastClient fasthttp.Client = fastClient()

// hamming接口元数据
type ImgMetaData struct {
	EcList []EcMetaData `json:"ecList"`
	MpList []MpMetaData `json:"mpList"`
}

type HammingDataInterface struct {
	EcRecord EcMetaData `json:"ecRecord"`
	MpRecord MpMetaData `json:"mpRecord"`
}

type ImgDataReqRes struct {
	ImgList []EcMetaData `json:"list"`
}

type EcMetaData struct {
	ImageURL      string `json:"imageUrl,omitempty"`
	ShopNo        string `json:"shopNo"`
	ImageHashCode string `json:"imageHashCode,omitempty"`
	Score         string `json:"score,omitempty"`
	LeftOrRight   string `json:"leftOrRight"`
}

type MpMetaData struct {
	ShopNo        string `json:"shopNo"`
	ImageURL      string `json:"imageUrl,omitempty"`
	ImageHashCode string `json:"imageHashCode,omitempty"`
}

// hamming接口 返回结构
type HammingResp struct {
	List []HammingScore `json:"list"`
}

type HammingScore struct {
	ShopNo        string       `json:"shopNo"`
	ImageURL      string       `json:"imageUrl,omitempty"`
	ImageHashCode string       `json:"imageHashCode,omitempty"`
	MatchEcList   []EcMetaData `json:"ecList"`
}

type tmp struct {
	metaDataTmp          ImgDataReqRes
	failRequestNumber    int32
	successRequestNumber int32
}

// 生成图片hash
func GetImgHash(c *gin.Context) {
	var (
		appG     = app.Gin{C: c}
		metaData ImgDataReqRes
	)
	httpCode, errCode := app.BindAndValid(c, &metaData)
	if errCode != e.SUCCESS {
		appG.Response(httpCode, errCode, nil)
		return
	}
	t := new(tmp)
	goEcHash(metaData.ImgList, t)
	fmt.Printf("结束....,失败请求数==%d, 成功请求数==%d", t.failRequestNumber, t.successRequestNumber)
	appG.Response(http.StatusOK, e.SUCCESS, t.metaDataTmp)
}

// 获取图片相似度分数
func GetHammingScore(c *gin.Context) {
	var (
		appG     = app.Gin{C: c}
		metaData ImgMetaData
	)
	httpCode, errCode := app.BindAndValid(c, &metaData)
	if errCode != e.SUCCESS {
		appG.Response(httpCode, errCode, nil)
		return
	}
	respDataList := new(HammingResp)
	defer ants.Release()
	var wg sync.WaitGroup
	p, err := ants.NewPoolWithFunc(100, func(v interface{}) {
		interfaceHamming(v, respDataList)
		// time.Sleep(time.Duration(100) * time.Millisecond)
		wg.Done()
	})
	if err != nil {
		return
	}
	if p == nil {
		log.Printf("协程池创建失败==%s", err)
		return
	}
	defer p.Release()

	tmpData := HammingDataInterface{}
	for _, record := range metaData.MpList {
		if len(record.ImageHashCode) == 0 {
			continue
		}
		for _, ecRecord := range metaData.EcList {
			if len(ecRecord.ImageHashCode) == 0 {
				continue
			}
			if record.ShopNo != ecRecord.ShopNo {
				continue
			}
			tmpData.MpRecord = record
			tmpData.EcRecord = ecRecord
			wg.Add(1)
			if err = p.Invoke(tmpData); err != nil {
				continue
			}
		}
	}
	wg.Wait()
	appG.Response(http.StatusOK, e.SUCCESS, respDataList)

}

func fastClient() fasthttp.Client {
	return fasthttp.Client{
		Name:                     "FunTester",
		NoDefaultUserAgentHeader: true,
		TLSConfig:                &tls.Config{InsecureSkipVerify: true},
		MaxConnsPerHost:          2000,
		MaxIdleConnDuration:      1 * time.Second,
		MaxConnDuration:          1 * time.Second,
		ReadTimeout:              1 * time.Second,
		WriteTimeout:             1 * time.Second,
		MaxConnWaitTimeout:       1 * time.Second,
	}
}

// fastGet 获取GET请求对象,没有进行资源回收
// @Description:
// @param url
// @param args
// @return *fasthttp.Request
func fastGet(url string) *fasthttp.Request {
	req := fasthttp.AcquireRequest()
	req.Header.SetMethod("GET")
	req.SetRequestURI(url)
	return req
}

// FastResponse 获取响应,保证资源回收
// @Description:
// @param request
// @return []byte
// @return error
func fastResponse(request *fasthttp.Request) ([]byte, error) {
	response := fasthttp.AcquireResponse()
	defer fasthttp.ReleaseRequest(request)
	defer fasthttp.ReleaseResponse(response)
	if err := FastClient.Do(request, response); err != nil {
		log.Printf("响应出错了==%s", err)
		return nil, err
	}
	return response.Body(), nil
}

func getImg(url string) (image.Image, error) {
	get := fastGet(url)
	byteBody, _ := fastResponse(get)
	reader := bytes.NewReader(byteBody)
	img, _, err := imageorient.Decode(reader)
	return img, err
}

func interfaceHamming(v interface{}, respDataList *HammingResp) {
	data := v.(HammingDataInterface)
	hash1, _ := goimagehash.ExtImageHashFromString(data.MpRecord.ImageHashCode)
	hash2, _ := goimagehash.ExtImageHashFromString(data.EcRecord.ImageHashCode)
	distance, _ := hash1.Distance(hash2)
	// 找hamming 差距小于2的 {0,1,2}
	if distance < 2 {
		res := &HammingScore{}
		res.ImageURL = data.MpRecord.ImageURL
		res.ImageHashCode = data.MpRecord.ImageHashCode
		res.ShopNo = data.MpRecord.ShopNo
		data.EcRecord.Score = strconv.Itoa(distance)
		res.MatchEcList = append(res.MatchEcList, data.EcRecord)
		respDataList.List = append(respDataList.List, *res)
	}

}

func doHash(v interface{}, t *tmp) (err error) {
	if v == nil || v == "" {
		t.failRequestNumber++
		fmt.Println("未指定图片URL")
		return errors.New("未指定图片URL")
	}
	url := v.(EcMetaData)
	img, err := getImg(url.ImageURL)
	if err != nil {
		fmt.Println(err.Error())
		t.failRequestNumber++
		return
	}
	if img == nil {
		fmt.Println(err.Error())
		t.failRequestNumber++
		return errors.New("获取图片失败")
	}
	hash1, err := goimagehash.PerceptionHash(img)
	if err != nil {
		fmt.Println(err.Error())
		t.failRequestNumber++
		return err //这个err是局部变量，非方法返回值定义的err所以需要return指定变量
	}
	if hash1 == nil {
		fmt.Println(err.Error())
		t.failRequestNumber++
		return errors.New("图片哈希失败")
	}
	url.ImageHashCode = hash1.ToString()
	t.metaDataTmp.ImgList = append(t.metaDataTmp.ImgList, url)
	t.successRequestNumber++
	return
}

// 最终返回list
func goEcHash(list []EcMetaData, t *tmp) (err error) {
	defer ants.Release()
	var wg sync.WaitGroup
	beg := time.Now()
	p, err := ants.NewPoolWithFunc(routineCountTotal, func(v interface{}) {
		doHash(v, t)
		wg.Done()
	})
	if err != nil {
		return
	}
	if p == nil {
		return errors.New("协程池创建失败")
	}
	defer p.Release()
	for i := 0; i < len(list); i++ {
		wg.Add(1)
		if err = p.Invoke(list[i]); err != nil {
			continue
		}
	}
	wg.Wait()
	fmt.Printf("time consumed: %fs", time.Now().Sub(beg).Seconds())
	return
}

图片相似度比对 ​

一、Web Worker方式实现 -- 浏览器 ​

Ⅰ. pixelmatch + thread.js ​

1. 依赖安装与配置 ​

2. 业务逻辑 ​

main线程 ​

worker线程 ​

Ⅱ. resemble + thread.js ​

1. 依赖安装与配置 ​

Resemble.js -- 修改版 ​

2. 业务逻辑 ​

**main线程 ** ​

worker线程 ​

性能测试 ​

多核与性能的关系 ​

结论 ​

二、java图片相似度计算 ​

三、nodejs 图片相似度计算 ​

依赖库 ​

项目结构目录 ​

实现 ​

1. app配置router ​

2. routes：image-diff.js ​

3. services：ImgPiscinaService.js ​

4. workers: worker.js ​

性能 ​

四、golang图片相似度计算 ​

ImgService.go ​

图片相似度比对

一、Web Worker方式实现 -- 浏览器

Ⅰ. pixelmatch + thread.js

1. 依赖安装与配置

2. 业务逻辑

main线程

worker线程

Ⅱ. resemble + thread.js

1. 依赖安装与配置

Resemble.js -- 修改版

2. 业务逻辑

main线程

worker线程

性能测试

多核与性能的关系

结论

二、java图片相似度计算

三、nodejs 图片相似度计算

依赖库

项目结构目录

实现

1. app配置router

2. routes：image-diff.js

3. services：ImgPiscinaService.js

4. workers: worker.js

性能

四、golang图片相似度计算

ImgService.go