OpenCV: cvFilter2D() 연습 코드

http://en.wikipedia.org/wiki/Convolution

void cvFilter2D(const CvArr* src, CvArr* dst, const CvMat* kernel, CvPoint anchor=cvPoint(-1, -1))¶

Convolves an image with the kernel.

Parameters:

src – The source image dst – The destination image kernel – Convolution kernel, a single-channel floating point matrix. If you want to apply different kernels to different channels, split the image into separate color planes using Split and process them individually anchor – The anchor of the kernel that indicates the relative position of a filtered point within the kernel. The anchor shoud lie within the kernel. The special default value (-1,-1) means that it is at the kernel center

elt3470@naver: 사용자가 kernel에 원하는 행렬을 입력함으로써, LPF, HPF 등을 직접 디자인해서 사용할 수 있습니다.

=> 그러므로, DoG (Derivative of Gaussian) 필터도 만들어 넣을 수 있다.



예로, 5x5 Gaussian  kernel을 만들어서 필터링하면 다음과 같이 영상을 smoothing하게 된다.

ref. 
2010/04/03 - [Visual Information Processing Lab] - Image Filtering

/* Test: 2-D DOG (Derivative of Gaussian) filter using cvFilter2D() in OpenCV
 ref.
 1) http://opencv.willowgarage.com/documentation/image_filtering.html?highlight=cvsobel#cvFilter2D
 2) Richard Szeliski, Computer Vision: Algorithms and Applications 115쪽
 camera: Logitech QuickCam Pro 4000
 */

#include <OpenCV/OpenCV.h>
#include <iostream>
using namespace std;

int main()
{
    IplImage* iplInput = 0; // input image
    IplImage* iplGray = 0; // grey image converted from input image
    IplImage* iplGauss = 0; // filtered image by 5x5 Gaussian kernel
    
    // define a kernel
    double gaussian[] = { 1, 4, 6, 4, 1,
                          2, 8, 12, 8, 2,
                          6, 24, 36, 24, 6,
                          2, 8, 12, 8, 2,
                          1, 4, 6, 3, 1 };
    for( int k = 0; k < 25; k++ )
    {
        gaussian[k] = gaussian[k] / 256;
    }
    CvMat G = cvMat( 5, 5, CV_64FC1, gaussian );
    
    char title[200];
    sprintf(title, "filtered image by Gaussian kernel");    
    
    cvNamedWindow("input");
    cvNamedWindow(title);
    
    // initialize capture from a camera
    CvCapture* capture = cvCaptureFromCAM(0); // capture from video device #0
    int count = 0; // number of grabbed frames
    
    while(1) {
        // get video frames from the camera
        if ( !cvGrabFrame(capture) ) {
            printf("Could not grab a frame\n\7");
            exit(0);
        }
        else {
            cvGrabFrame( capture ); // capture a frame
            iplInput = cvRetrieveFrame(capture); // retrieve the caputred frame
           
            if(iplInput) {
                if(0 == count) {       
                    // create an image header and allocate the image data
                    iplGray = cvCreateImage(cvGetSize(iplInput), 8, 1);   
                    iplGauss = cvCreateImage(cvGetSize(iplInput), 8, 1);   
                }
                // convert the input color image to gray one
                cvCvtColor(iplInput, iplGray, CV_BGR2GRAY); // convert an image from one color space to another  
                count++;
            }           
            cvShowImage( "input", iplInput );
           
            // convolve an image with the kernel
            // void cvFilter2D(const CvArr* src, CvArr* dst, const CvMat* kernel, CvPoint  anchor=cvPoint(-1, -1)
            cvFilter2D( iplGray, iplGauss, &G ); // convolve an image with 5x5 Gaussian kernel        
              cvShowImage( title, iplGauss );
           
            // save images to files
            cvSaveImage("input.bmp", iplInput);
            cvSaveImage("gray.bmp", iplGray);
            cvSaveImage("DoG.bmp", iplGauss);
           
            if( cvWaitKey(10) >= 0 )
                break;
        }
    }   
    
    cvReleaseCapture( &capture ); // release the capture source
    cvDestroyWindow("input");
    cvDestroyWindow(title);
    
    return 0;   
}

입력 영상

흑백 영상

2차원 5x5 Gaussian convolution 영상 (smoothing)