OpenCV resizing and cropping image according to pixel value

Question

#include "iostream"
#include "cv.h"
#include "highgui.h"
#include "cvaux.h"
#include "cxmisc.h"
#include "math.h"

using namespace cv;
using namespace std;

int main(){

int height, width, x, y, i, minX, minY, maxX, maxY;
char imgFileName[100];


IplImage *origImage = cvLoadImage("BaybayinMark/b9.jpg", -1);
height = origImage->height;
width = origImage->width;

IplImage *grayImage = cvCreateImage(cvSize(width, height), 8, 1);
IplImage *binImage = cvCreateImage(cvSize(width, height), 8, 1);


//Pre-processing phase


cvCvtColor(origImage, grayImage, CV_BGR2GRAY);
cvDilate(grayImage, grayImage, NULL, 1);
cvSmooth(grayImage, grayImage, CV_GAUSSIAN, 21, 21, 0, 0);
cvThreshold(grayImage, binImage, 120, 255, CV_THRESH_BINARY);
cvNormalize(binImage,binImage,0,1,CV_MINMAX);

minX = width;
minY = height;
maxX = 0;
maxY = 0;


CvScalar s;


for (x=0; x<width-1; x++){
for(y=0; y<height-1; y++){
    s = cvGet2D(binImage, y, x);
    //printf("%f\n", s.val[0]);
    if (s.val[0] == 1){
        //printf("HELLO");
        minX = min(minX, x);
        minY = min(minY, y);
        maxX = max(maxX, x);
        maxY = max(maxY, y);

    }   
}
}

cvSetImageROI(binImage, cvRect(minX, minY, maxX-minX, maxY-minY));

IplImage *cropImage = cvCreateImage(cvGetSize(binImage), 8, 1);

cvCopy(binImage, cropImage, NULL);

cvSaveImage("crop/cropImage9.jpg", cropImage);
cvResetImageROI(binImage);

cvReleaseImage(&origImage);
cvReleaseImage(&binImage);

cvReleaseImage(&grayImage);
cvReleaseImage(&cropImage);

}

Hi! i just want to ask about this code. I am trying to identify the outermost edges of an image and crop the image according them. All I was having after running was a black image with the same size. Am I trying to do it the wrong way? Please enlighten me I'm a beginner with OpenCV.

I would just stick a bunch of `cvSaveImage()` calls in between of each step to see where it actually becomes incorrect. — dragonroot, Nov 04 '11 at 08:46
I did some editing. What might be the problem why the output image is just the same binary image? — cmsl, Nov 04 '11 at 14:25
Wasn't this question answered successfully? Please review the answers and accept the one that solved the problem. Simply click on the checkbox near it to select the official answer. By doing this you will be helping future visitors. — karlphillip, Feb 05 '13 at 13:21

score 12 · Answer 1 · edited May 22 '15 at 21:09

In the rush of finding what-the-heck-is-the-problem people tend to forget a more important question: how-the-heck-do-i-find-the-problem.

With image processing applications, the how can be answered by the poor man's debugger in OpenCV, which is adding cvSaveImage() calls through the code to be able to visualize what every step of the way is doing:

    //Pre-processing phase
    cvCvtColor(origImage, grayImage, CV_BGR2GRAY);
    cvSaveImage("cv_color.jpg", grayImage);

    cvDilate(grayImage, grayImage, NULL, 1);
    cvSaveImage("cv_dilate.jpg", grayImage);

    cvSmooth(grayImage, grayImage, CV_GAUSSIAN, 21, 21, 0, 0);
    cvSaveImage("cv_smooth.jpg", grayImage);

    cvThreshold(grayImage, binImage, 120, 255, CV_THRESH_BINARY);
    cvSaveImage("cv_threshold.jpg", binImage);

    cvNormalize(binImage,binImage,0,1,CV_MINMAX);
    cvSaveImage("cv_normalize.jpg", binImage);

This code reveals that the resulting image gets black even before your custom for loop, and the call that is responsible for that is cvNormalize(). But it makes sense right? You are converting pixels which are in the range [0..255] to values of 0 and 1.

So the problem is that at the end of your processing, when you save the resulting image to the dislk, you forgot to normalize the values back to the original range:

    IplImage *cropImage = cvCreateImage(cvGetSize(binImage), 8, 1);
    cvCopy(binImage, cropImage, NULL);

    cvNormalize(cropImage, cropImage, 0, 255, CV_MINMAX);

    cvSaveImage("result.jpg", cropImage);

And that solves the problem.

OpenCV resizing and cropping image according to pixel value

1 Answers1

Linked