JAVA+OpenCV图像增强——直方图均衡、自适应直方图均衡、拉普拉斯局部增强、对数变换增强、伽马增强

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/u012525096/article/details/88135284

图像增强

图片太亮、太暗等问题。
环境：JAVA+Opencv4

直方图均衡

    /**
     * 增强对比度
     *
     * @param src BGR格式图像
     * @return
     */
    public static Mat histEqualize(Mat src) {
        Mat dst = src.clone();
        Imgproc.cvtColor(dst, dst, Imgproc.COLOR_BGR2YCrCb);
        List<Mat> list1 = new ArrayList<>();
        Core.split(dst, list1);
        Imgproc.equalizeHist(list1.get(0), list1.get(0));
        Core.normalize(list1.get(0), list1.get(0), 0, 255, Core.NORM_MINMAX);
        Core.merge(list1, dst);
        Imgproc.cvtColor(dst, dst, Imgproc.COLOR_YCrCb2BGR);
        return dst;
    }

对比度受限直方图均衡化CLAHE

    public static Mat autoHistEqualize(Mat src) {
        Mat dst = src.clone();
        Imgproc.cvtColor(dst, dst, Imgproc.COLOR_BGR2YCrCb);
        List<Mat> list1 = new ArrayList<>();
        Core.split(dst, list1);
        CLAHE clahe = Imgproc.createCLAHE();
        clahe.setClipLimit(4);
        clahe.apply(list1.get(0), list1.get(0));
//        Core.normalize(list1.get(0),list1.get(0),0,255,Core.NORM_MINMAX);
        Core.merge(list1, dst);
        Imgproc.cvtColor(dst, dst, Imgproc.COLOR_YCrCb2BGR);
        return dst;
    }

拉普拉斯算子增强

    /**
     * 用于整体偏暗图像的增强,变亮
     *
     * @param src
     * @return
     */
    public static Mat laplaceEnhance(Mat src) {
        Mat srcClone = src.clone();
        float[] kernel = {0, 0, 0, -1, 5f, -1, 0, 0, 0};
        Mat kernelMat = new Mat(3, 3, CvType.CV_32FC1);
        kernelMat.put(0, 0, kernel);
        Imgproc.filter2D(srcClone, srcClone, CvType.CV_8UC3, kernelMat);
        return srcClone;
    }

对数变换

变换使得暗部映射到大多数的范围，强调图像低灰度值的部分。

    /**
     * 对数变换可以将图像的低灰度值部分扩展，显示出低灰度部分更多的细节，
     * 将其高灰度值部分压缩，减少高灰度值部分的细节，从而达到强调图像低灰度部分的目的。
     *
     * @param src
     * @return
     */
    public static Mat logEnhance(Mat src) {
        Mat srcClone = src.clone();
        Mat imageResult = new Mat(srcClone.size(), CvType.CV_32FC3);
        Core.add(srcClone, new Scalar(5, 5, 5), srcClone);
        srcClone.convertTo(srcClone, CvType.CV_32F);
        Core.log(srcClone, imageResult);
//        Core.multiply(imageLog, new Scalar(3,3,3), imageLog);
        Core.normalize(imageResult, imageResult, 0, 255, Core.NORM_MINMAX);
        Core.convertScaleAbs(imageResult, imageResult);
        return imageResult;
    }

伽马变换

原理同上，γ大于1可强调高灰度值，γ小于1可强调低灰度值。

    public static Mat gammaEnhance(Mat src) {
        Mat srcClone = src.clone();
        srcClone.convertTo(srcClone, CvType.CV_32F);

        Core.pow(srcClone, 4, srcClone);

        Core.normalize(srcClone, srcClone, 0, 255, Core.NORM_MINMAX);
        Core.convertScaleAbs(srcClone, srcClone);

        return srcClone;
    }

Test

public class ImageEnhance {
    static {
        System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
    }

    public static void main(String[] args) {
        Mat dark, light, nopingheng, piandark, src;
        src = Imgcodecs.imread("images/1.png");
        dark = Imgcodecs.imread("images/dark.jpg");
        light = Imgcodecs.imread("images/light.jpg");
        nopingheng = Imgcodecs.imread("images/nopingheng.jpg");
        piandark = Imgcodecs.imread("images/piandark.jpg");

//        Imgcodecs.imwrite("result/1-histEqualize.png", histEqualize(src));
//        Imgcodecs.imwrite("result/dark-histEqualize.png", histEqualize(dark));
//        Imgcodecs.imwrite("result/light-histEqualize.png", histEqualize(light));
//        Imgcodecs.imwrite("result/nopingheng-histEqualize.png", histEqualize(nopingheng));
//        Imgcodecs.imwrite("result/piandark-histEqualize.png", histEqualize(piandark));
//
//        Imgcodecs.imwrite("result/1-autoHistEqualize.png", autoHistEqualize(src));
//        Imgcodecs.imwrite("result/dark-autoHistEqualize.png", autoHistEqualize(dark));
//        Imgcodecs.imwrite("result/light-autoHistEqualize.png", autoHistEqualize(light));
//        Imgcodecs.imwrite("result/nopingheng-autoHistEqualize.png", autoHistEqualize(nopingheng));
//        Imgcodecs.imwrite("result/piandark-autoHistEqualize.png", autoHistEqualize(piandark));
//
//        Imgcodecs.imwrite("result/1-laplaceEnhance.png", laplaceEnhance(src));
//        Imgcodecs.imwrite("result/dark-laplaceEnhance.png", laplaceEnhance(dark));
//        Imgcodecs.imwrite("result/light-laplaceEnhance.png", laplaceEnhance(light));
//        Imgcodecs.imwrite("result/nopingheng-laplaceEnhance.png", laplaceEnhance(nopingheng));
//        Imgcodecs.imwrite("result/piandark-laplaceEnhance.png", laplaceEnhance(piandark));
//
//        Imgcodecs.imwrite("result/1-logEnchane.png", logEnhance(src));
//        Imgcodecs.imwrite("result/dark-logEnchane.png", logEnhance(dark));
//        Imgcodecs.imwrite("result/light-logEnchane.png", logEnhance(light));
//        Imgcodecs.imwrite("result/nopingheng-logEnchane.png", logEnhance(nopingheng));
//        Imgcodecs.imwrite("result/piandark-logEnchane.png", logEnhance(piandark));
//
//        Imgcodecs.imwrite("result/1-gammaEnhance.png", gammaEnhance(src));
//        Imgcodecs.imwrite("result/dark-gammaEnhance.png", gammaEnhance(dark));
//        Imgcodecs.imwrite("result/light-gammaEnhance.png", gammaEnhance(light));
//        Imgcodecs.imwrite("result/nopingheng-gammaEnhance.png", gammaEnhance(nopingheng));
//        Imgcodecs.imwrite("result/piandark-gammaEnhance.png", gammaEnhance(piandark));
    }
}