如何使用ciexyz / cielab检测肤色?

时间:2017-04-26 15:27:24

标签: java opencv image-processing

我是opencv的新手,我一直在使用CIEXYZ进行皮肤检测。但我得到了将RGB转换为CIE Lab以获得肤色区域的问题,我根据this从RGB做了一些计算。

原始图片

enter image description here

结果是没有坚果黑框。它从RGB到CIEXYZ enter image description here

这是二进制图像

enter image description here

但我希望像这样显示它

enter image description here

这是我的源代码:

Mat img_color_space = new Mat();
Mat mask = new Mat();

Imgproc.cvtColor(src, img_color_space, colorBgr2hsv);
Imgcodecs.imwrite(path+"CIELAB/hsv.jpg",img_color_space);
Imgproc.blur(img_color_space, img_color_space, new Size(3,3));
Mat canny_output = new Mat();

Scalar minValues = new Scalar(0,10,60);
Scalar maxValues = new Scalar(20,150,255);
// show the current selected HSV range
String valuesToPrint = "Hue range: " + minValues.val[0] + "-" + maxValues.val[0]
        + "\tSaturation range: " + minValues.val[1] + "-" + maxValues.val[1] + "\tValue range: "
        + minValues.val[2] + "-" + maxValues.val[2];
//System.out.println("tresholding:"+valuesToPrint);

Core.inRange(img_color_space, minValues, maxValues, mask);
Imgcodecs.imwrite(path+"CIELAB/mask.jpg",mask);
List<MatOfPoint> contours = new ArrayList<>();
Mat hierarchy = new Mat();

Imgproc.findContours(mask, contours, hierarchy, Imgproc.RETR_TREE, Imgproc.CHAIN_APPROX_SIMPLE,new Point(0,0));
int s = findBiggestContour(contours);

Mat drawing = Mat.zeros(mask.size(), CvType.CV_8UC3);
Imgproc.drawContours(drawing, contours, s, new Scalar(255, 255, 255), -1,8,hierarchy,0,new Point(0,0));
Imgproc.blur(drawing, drawing, new Size(3,3));
Imgcodecs.imwrite(path+"CIELAB/biggest.jpg",drawing);

我的代码中有什么问题吗?提前谢谢!

1 个答案:

答案 0 :(得分:1)

您可以更简单的方式对手进行分段。

按原样读取CIELAB图像并将其分成三个不同的通道。分别分析每个通道,看看哪一个最佳分段。之后应用阈值。

以下代码在python中,可以转换为java:

import cv2

filename = 'hand.jpg'
img = cv2.imread(filename)
blue_channel, green_channel, red_channel = cv2.split(img)
cv2.imshow('green_channel', green_channel)

这是图片的绿色通道:

enter image description here

#---I split the image in blue, green and red channels because the image I saved is in BGR format  ---#

#---I applied binary threshold to the green channel---#
ret, thresh = cv2.threshold(g, 152, 255, 1)
cv2.imshow('thresh', thresh)
#--- I got the following----#

enter image description here

现在你可以找到最大的轮廓并单独划分手