OPENCV图像边缘查找与分割技术在android中使用汇总

图像分割技术或者叫抠图技术，是一种根据需要对图像进行截取分离的技术，在一般的图像处理和视频处理中应用十分广泛，是图像查找，图像识别，图像特效的基础。经常被人们使用在相机美颜，自动人脸马赛克，车牌识别，图像查找，人脸查找，人脸识别，机器视觉，AR等领域。

图像分割分有标注和无标注两种情况，一种是自动根据分割，自选阀值，区域自动分割，一种是在给定条件下分割，比如分割人脸，人身体，给定区域分割，前一种由于准确度的问题，应用很少。

主要图像分割方法包括阈值分割，边缘分割（查找边缘），区域分割（种子区域生长法、区域分裂合并法和分水岭法等），图论分割，能量泛函（参数活动轮廓模型，几何活动轮廓模型）等。利用机器学习自动对像素分类也能达到某些分割 目的。

OPENCV封装的分割算法非常多，而且又能根据需要组合使用，以提升分割 精度，这使得用法灵活性大增，掌握的难度比较 大。比较重要的有阀值分割Imgproc.threshold， 边缘分割Imgproc.findContours，也可以使用Roberts 算子、Laplace 算子、Prewitt 算子、Sobel 算子、Rosonfeld算子、Kirsch 算子以及Canny 算子等实现，区域分割HSV亮度， 图论分割GraphCut，GrabCut和Random Walk，还可以根据颜色分割等。有些精度高，耗时长，有些分割粗糙，但速度快，需要根据需要自由组合使用：

import java.util.ArrayList;
import java.util.List;

import org.opencv.core.Core;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
import org.opencv.core.MatOfPoint;
import org.opencv.core.MatOfPoint2f;
import org.opencv.core.Point;
import org.opencv.core.Rect;
import org.opencv.core.Scalar;
import org.opencv.core.Size;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;

public class ImageOpencv {
	public static void main(String[] args) {
		System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
		Mat src = Imgcodecs.imread("E:/work/qqq/a9.jpg");
    	Imgcodecs.imwrite("E:/work/qqq/hh1.jpg", removeBackground(src));
    	
    	Mat src2 = Imgcodecs.imread("E:/work/qqq/a3.jpg");
    	Imgcodecs.imwrite("E:/work/qqq/hh2.jpg", MyThresholdHsv(src2));
    	
    	
    	Mat src3 = Imgcodecs.imread("E:/work/qqq/e1.jpg");
        Imgcodecs.imwrite("E:/work/qqq/hh3.jpg", myGrabCut(src3,  new Point(50,0),new Point(300, 250)));
    	
        Mat src4 = Imgcodecs.imread("E:/work/qqq/dd.jpg");
        Imgcodecs.imwrite("E:/work/qqq/hh4.jpg", MyFindLargestRectangle(src4));
        
        Mat src5 = Imgcodecs.imread("E:/work/qqq/dd.jpg");
        Imgcodecs.imwrite("E:/work/qqq/hh5.jpg", MyWatershed(src5));
       
        Mat src6 = Imgcodecs.imread("E:/work/qqq/e1.jpg");
        Imgcodecs.imwrite("E:/work/qqq/hh6.jpg", MyCanny(src6, 100));
        
    	SkinDetection sd= new SkinDetection(Imgcodecs.imread("E:/work/qqq/e1.jpg"));
    	Imgcodecs.imwrite("E:/work/qqq/hh7.jpg",sd.GetSkin());   	

     Mat src7 = Imgcodecs.imread("E:/work/qqq/ee.jpg");
     Imgcodecs.imwrite("E:/work/qqq/hh8.jpg", MyFloodFill(src7));
	}

	// threshold根据反差去掉深色单色背景
	public static Mat removeBackground(Mat nat) {
		Mat m = new Mat();

		Imgproc.cvtColor(nat, m, Imgproc.COLOR_BGR2GRAY);
		double threshold = Imgproc.threshold(m, m, 0, 255, Imgproc.THRESH_OTSU);
		Mat pre = new Mat(nat.size(), CvType.CV_8UC3, new Scalar(0, 0, 0));
		Mat fin = new Mat(nat.size(), CvType.CV_8UC3, new Scalar(0, 0, 0));
		for (int i = 0; i < m.rows(); i++) {
			for (int j = 0; j < m.cols(); j++) {
				double[] ds = m.get(i, j);
				double[] data = { ds[0] / 255, ds[0] / 255, ds[0] / 255 };
				pre.put(i, j, data);
			}
		}
		for (int i = 0; i < pre.rows(); i++) {
			for (int j = 0; j < pre.cols(); j++) {
				double[] pre_ds = pre.get(i, j);
				double[] nat_ds = nat.get(i, j);
				double[] data = { pre_ds[0] * nat_ds[0], pre_ds[1] * nat_ds[1], pre_ds[2] * nat_ds[2] };
				fin.put(i, j, data);
			}
		}

		return fin;
	}

	// threshold根据亮度去除背景
	private static Mat MyThresholdHsv(Mat frame) {
		Mat hsvImg = new Mat();
		List<Mat> hsvPlanes = new ArrayList<>();
		Mat thresholdImg = new Mat();

		// threshold the image with the average hue value
		hsvImg.create(frame.size(), CvType.CV_8U);
		Imgproc.cvtColor(frame, hsvImg, Imgproc.COLOR_BGR2HSV);
		Core.split(hsvImg, hsvPlanes);
		// get the average hue value of the image
		Scalar average = Core.mean(hsvPlanes.get(0));
		double threshValue = average.val[0];
		Imgproc.threshold(hsvPlanes.get(0), thresholdImg, threshValue, 179.0, Imgproc.THRESH_BINARY_INV);

		Imgproc.blur(thresholdImg, thresholdImg, new Size(15, 15));

		// dilate to fill gaps, erode to smooth edges
		Imgproc.dilate(thresholdImg, thresholdImg, new Mat(), new Point(-1, -1), 1);
		Imgproc.erode(thresholdImg, thresholdImg, new Mat(), new Point(-1, -1), 3);

		Imgproc.threshold(thresholdImg, thresholdImg, threshValue, 179.0, Imgproc.THRESH_BINARY);

		// create the new image
		Mat foreground = new Mat(frame.size(), CvType.CV_8UC3, new Scalar(0, 0, 0));
		thresholdImg.convertTo(thresholdImg, CvType.CV_8U);
		frame.copyTo(foreground, thresholdImg);
		return foreground;
	}
    //grabCut分割技术
	public static Mat myGrabCut(Mat in, Point tl, Point br) {
		Mat mask = new Mat();
		Mat image = in;
		mask.create(image.size(), CvType.CV_8UC1);
		mask.setTo(new Scalar(0));

		Mat bgdModel = new Mat();// Mat.eye(1, 13 * 5, CvType.CV_64FC1);
		Mat fgdModel = new Mat();// Mat.eye(1, 13 * 5, CvType.CV_64FC1);

		Mat source = new Mat(1, 1, CvType.CV_8U, new Scalar(3));
		Rect rectangle = new Rect(tl, br);
		Imgproc.grabCut(image, mask, rectangle, bgdModel, fgdModel, 3, Imgproc.GC_INIT_WITH_RECT);
		Core.compare(mask, source, mask, Core.CMP_EQ);
		Mat foreground = new Mat(image.size(), CvType.CV_8UC1, new Scalar(0, 0, 0));
		image.copyTo(foreground, mask);

		return foreground;

	}
	//findContours分割技术
	private static Mat MyFindLargestRectangle(Mat original_image) {
		Mat imgSource = original_image;
		Imgproc.cvtColor(imgSource, imgSource, Imgproc.COLOR_BGR2GRAY);
		Imgproc.Canny(imgSource, imgSource, 50, 50);
		Imgproc.GaussianBlur(imgSource, imgSource, new Size(5, 5), 5);
		List<MatOfPoint> contours = new ArrayList<MatOfPoint>();
		Imgproc.findContours(imgSource, contours, new Mat(), Imgproc.RETR_LIST, Imgproc.CHAIN_APPROX_SIMPLE);
		double maxArea = 0;
		int maxAreaIdx = -1;
		MatOfPoint largest_contour = contours.get(0); 
		MatOfPoint2f approxCurve = new MatOfPoint2f();
		for (int idx = 0; idx < contours.size(); idx++) {
			MatOfPoint temp_contour = contours.get(idx);
			double contourarea = Imgproc.contourArea(temp_contour);
			if (contourarea - maxArea > 1) {
				maxArea = contourarea;
				largest_contour = temp_contour;
				maxAreaIdx = idx;
				MatOfPoint2f new_mat = new MatOfPoint2f(temp_contour.toArray());
				int contourSize = (int) temp_contour.total();
				Imgproc.approxPolyDP(new_mat, approxCurve, contourSize * 0.05, true);
			}
		}


		Imgproc.drawContours(imgSource, contours, -1, new Scalar(255, 0, 0), 1);
		Imgproc.fillConvexPoly(imgSource, largest_contour, new Scalar(255, 255, 255));
		Imgproc.drawContours(imgSource, contours, maxAreaIdx, new Scalar(0, 0, 255), 3);

		return imgSource;
	}
	//watershed分水岭分割技术
    public static Mat MyWatershed(Mat img)
    {
        Mat threeChannel = new Mat();

        Imgproc.cvtColor(img, threeChannel, Imgproc.COLOR_BGR2GRAY);
        //Imgproc.threshold(threeChannel, threeChannel, 200, 255, Imgproc.THRESH_BINARY);
        Imgproc.threshold(threeChannel, threeChannel, 0, 255, Imgproc.THRESH_OTSU);
        
        Mat fg = new Mat(img.size(),CvType.CV_8U);
        Imgproc.erode(threeChannel,fg,new Mat());

        Mat bg = new Mat(img.size(),CvType.CV_8U);
        Imgproc.dilate(threeChannel,bg,new Mat());
        Imgproc.threshold(bg,bg,1, 128,Imgproc.THRESH_BINARY_INV);

        Mat markers = new Mat(img.size(),CvType.CV_8U, new Scalar(0));
        Core.add(fg, bg, markers);

        Mat result=new Mat();
        markers.convertTo(result, CvType.CV_32SC1);
        Imgproc.watershed(img, result);
        result.convertTo(result,CvType.CV_8U);
        
        
        return result;
    }
    //Canny分割技术
    public static Mat MyCanny(Mat img, int threshold) {
        Imgproc.cvtColor(img, img, Imgproc.COLOR_BGR2GRAY);
        Imgproc.Canny(img, img, threshold, threshold * 3, 3, true);
        return img;
    }   
//漫水填充
 public static Mat MyFloodFill(Mat img) {
        Rect ccomp = new Rect();
        Mat mask = new Mat();
        Imgproc.floodFill(img, mask, new Point(50,20), new Scalar(0, 0,0), ccomp,new Scalar(10, 10, 10), new Scalar(10, 10, 10), 0);
        
        return img;
    }  

}

上图：

removeBackground，较深单色背景时

MyThresholdHsv，较浅亮色背景

myGrabCut去除复杂背景

查找边缘轮廓，提取mask用于分割

MyWatershed

MyCanny和基于肤色分割

参考 ：

http://blog.csdn.net/zouxy09/article/details/8532106
http://blog.csdn.net/vast_sea/article/details/8196507
http://blog.csdn.net/bless2015/article/details/52805875
http://www.07net01.com/2016/04/1458778.html

http://blog.csdn.net/dcrmg/article/details/52498440

http://blog.csdn.net/liang_dun/article/details/45198911