CODEWITHSUNDEEP
c++opencv
hettie
hettie

Reputation: 43

OpenCV: Applying Operations on RGB Images (Split+Merge)

Sorry if this question is so basic, I'm new with OpenCV.

I want to apply histogram equalisation to an RGB image. HE only works for single channel so I figured that I have to split the image into 3 different channels, apply HE to each one of them, then merge them all together to form output equalised image.

So I did just that and here's my program which I REALLY believe should work:

#include <iostream>
#include <stdio.h>
#include <cv.h>
#include <highgui.h>

int main(int argc, char * argv[])
{
IplImage* img = cvLoadImage("/Users/Documents/red.jpg");  //Load image file
cvNamedWindow("Trans-in", CV_WINDOW_AUTOSIZE );     //Create input window
cvNamedWindow("Trans-out", CV_WINDOW_AUTOSIZE  );   //Create output window

cvShowImage("Trans-in", img );  //Show input image in input window

//Create space for outputs rgb and its separate channels, r, g and b
IplImage* img0 = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 3);    //rgb
IplImage* r = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 1);    //r
IplImage* g = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 1);    //g
IplImage* b = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 1);    //b

//cvZero(r);
//cvZero(g);
//cvZero(b);

cvSplit(img, b, g, r, NULL);       //OpenCV likes it in BGR format

cvEqualizeHist( img, r );    //equalise r
cvEqualizeHist( img, g );    //equalise g
cvEqualizeHist( img, b );    //equalise b


cvMerge(b, g, r, NULL, img0);  //merge all separate channels together to output image rgb

//cvReleaseImage(&r);
//cvReleaseImage(&g);
//cvReleaseImage(&b);

cvShowImage("Trans-out", img0); //Show output image in output window


cvWaitKey(0);
cvReleaseImage( &img);
cvReleaseImage( &img0);

cvDestroyWindow( "Trans-in");
cvDestroyWindow( "Trans-out");

return 0;
}

Apologies if the //comments are a bit annoying, but it shows my 'reasoning' I guess. Xcode didn't complain until I hit the 'Run' button. It basically crashed (input image shows, but not output image).

Plus an error message at the bottom:

OpenCV Error: Assertion failed (CV_ARE_SIZES_EQ(src, dst) && CV_ARE_TYPES_EQ(src, dst) && CV_MAT_TYPE(src->type) == CV_8UC1) in cvEqualizeHist, file /opt/local/var/macports/build/_Volumes_work_mports_dports_graphics_opencv/opencv/work/OpenCV-2.4.3/modules/imgproc/src/histogram.cpp, line 2414 terminate called throwing an exception

And I have no idea what it means or what to do next.

Upvotes: 4

Views: 13691

Answers (2)

rotating_image
rotating_image

Reputation: 3076

For doing histogram equalization to a RGB image...convert it to HSV image and then equalize the V plane...

  1. convert RGB to HSV
  2. split HSV into H,S and V
  3. equalize V

  4. merge H,S and V

    vector<Mat> hsv_planes;
cvtColor(image,image,CV_BGR2HSV);
split(image,hsv_planes);
equalizeHist(hsv_planes[2],hsv_planes[2]);
merge(hsv_planes,image);
cvtColor(image,image,CV_HSV2BGR);

Upvotes: 2

sgarizvi
sgarizvi

Reputation: 16796

You are using the cvEqualizeHist function incorrectly in the following lines:

cvEqualizeHist( img, r );    //equalise r
cvEqualizeHist( img, g );    //equalise g
cvEqualizeHist( img, b );    //equalise b

img is a 3 channel image while r,g and b are single channel images. cvEqualizeHist doesn't work on 3 channel images.

You have to do the following:

cvEqualizeHist( r, r );    //equalise r
cvEqualizeHist( g, g );    //equalise g
cvEqualizeHist( b, b );    //equalise b

Upvotes: 4

Related Questions