Separable kernels from OpenCV

Hello guys,

I am learning OpenCV by the book "OpenCV...." and cannot make one exercise there:

*****************
Separable kernels. Create a 3-by-3 Gaussian kernel using rows [(1/16, 2/16, 1/16), (2/16, 4/16, 2/16), (1/16, 2/16, 1/16)] and with anchor point in the middle.

a. Run this kernel on an image and display the results.

b. Now create two one-dimensional kernels with anchors in the center: one going “across” (1/4, 2/4, 1/4), and one going down (1/4, 2/4, 1/4). Load the same original image and use cvFilter2D() to convolve the image twice, once with the first 1D kernel and once with the second 1D kernel. Describe the results.
************************

I did three those filters, but they do not change my source image, why, where is my error?

Here is my code:

Code:

---

                IplImage* src = cvLoadImage("4.jpg");
                cvNamedWindow("src", 1);
                cvShowImage("src", src);
                IplImage* rgb[3];
                float L45[]={
                        1./16., 2./16., 1./16.,
                        2./16., 4./16., 2./16.,
                        1./16., 2./16., 1./16.};// 
                CvMat*  rgbMat = cvCreateMat(3, 3, CV_32FC1);
                for (int y = 0; y < 3; y++)
                {
                        for (int x = 0; x < 3; x++)
                                cvmSet(rgbMat, y, x, L45[y*3 + x]);
                }
                IplImage* dst = cvCreateImage(cvSize(src->width, src->height), src->depth, 3);
                IplImage* dstRGB[3];
                for (int i = 0; i < 3; i++)
                {
                        rgb[i] = cvCreateImage(cvSize(src->width, src->height), src->depth, 1);
                        dstRGB[i] = cvCreateImage(cvSize(src->width, src->height), src->depth, 1);
                }
                cvSplit(src, rgb[0], rgb[1], rgb[2], NULL);
                for (int i = 0; i < 3; i++)
                {
                        cvFilter2D(rgb[i], dstRGB[i], rgbMat);
                }
                cvReleaseMat(&rgbMat);
                cvMerge(dstRGB[0], dstRGB[1], dstRGB[2], NULL, dst);
                cvNamedWindow("dst", 1);
                cvShowImage("dst", dst);
                //across
                float D[] = {1./4., 2./4., 1./4.};
                rgbMat = cvCreateMat(1, 3, CV_32FC1);
                for (int x = 0; x < 3; x++)
                        cvmSet(rgbMat, 0, x, D[x]);
                for (int i = 0; i < 3; i++)
                {
                        cvZero(dstRGB[i]);
                        cvFilter2D(rgb[i], dstRGB[i], rgbMat, cvPoint(1, 0));
                }
                cvReleaseMat(&rgbMat);
                cvZero(dst);
                cvMerge(dstRGB[0], dstRGB[1], dstRGB[2], NULL, dst);
                cvNamedWindow("Across", 1);
                cvShowImage("Across", dst);
                //down
                rgbMat = cvCreateMat(3, 1, CV_32FC1);
                for (int y = 0; y < 3; y++)
                        cvmSet(rgbMat, y, 0, D[y]);
                for (int i = 0; i < 3; i++)
                {
                        cvZero(dstRGB[i]);
                        cvFilter2D(rgb[i], dstRGB[i], rgbMat, cvPoint(0, 1));
                }
                cvReleaseMat(&rgbMat);
                cvZero(dst);
                cvMerge(dstRGB[0], dstRGB[1], dstRGB[2], NULL, dst);
                cvNamedWindow("Down", 1);
                cvShowImage("Down", dst);
                cvReleaseImage(&dst);
                for (int i = 0; i < 3; i++)
                {
                        cvReleaseImage(&rgb[i]);
                        cvReleaseImage(&dstRGB[i]);
                }
                cvWaitKey(0);
                cvReleaseImage(&src);
                cvDestroyWindow("src");
                cvDestroyWindow("dst");
                cvDestroyWindow("Across");
                cvDestroyWindow("Down");

---

I know nothing about your code, but I don't think that

Quote:

---

using rows [(1/16, 2/16, 1/16), (2/16, 4/16, 2/16), (1/16, 2/16, 1/16)]

---

is the same as :

Code:

---

float L45[]={
1./16., 2./16., 1./16.,
2./16., 4./16., 2./16.,
1./16., 2./16., 1./16.};

---

L45[0] == 1 divided by 16 = 0.0625.

... but I could be totally wrong ofcourse ;)

Quote:

---

Originally Posted by Skizmo

I know nothing about your code, but I don't think that



is the same as :

Code:

---

float L45[]={
1./16., 2./16., 1./16.,
2./16., 4./16., 2./16.,
1./16., 2./16., 1./16.};

---

L45[0] == 1 divided by 16 = 0.0625.

... but I could be totally wrong ofcourse ;)

---

The OpenCV requires float values for future handling , so 0.0625 is a right value.

Hello,

I ran your code... it seems to be working fine ! :)
I guess it is just that the kernel size is a bit small (3x3) that the change in between the source and destination images is not very noticeable.
In order to be sure I just did a difference, in your case just do something like:

Code:

---

  cvSub(src, dst, dst);
  cvNamedWindow("dst", 1);
  cvShowImage("dst", dst);

---

or use a different source image with some fine lines or noise etc...

Hope it helps, cya !

Convolving a Gaussian kernel with an image simply blurs it a bit.