OpenCV BGR值修改

Question

The following code is supposed to change all the values of BGR to 128.

#include<stdio.h>
#include<opencv2/opencv.hpp>
#include<iostream>
using namespace std;
using namespace cv;
int main(int argc,char** argv)
{

    Mat input_image=imread(argv[1],1);
    if(!input_image.data)
    {
        printf("No image data\n");
        return -1;
    }
    long nRow = input_image.rows;
    long nCol = input_image.cols;

    if(input_image.isContinuous())
    {
        nCol *= nRow;
        nRow = 1;
    }

    long i,j;
    uchar* p;
    for(i=0;i<nRow;i++)
    {
        p=input_image.ptr<uchar>(i);
        for(j=0;j<nCol;j++)
        {
            p[j]=128;
        }
    }
    namedWindow("Changed color",CV_WINDOW_AUTOSIZE);
    imshow("Changed color",input_image);

    waitKey(0);

    return 0;
}

The image name is given as a command line argument. This code is compiled using openCV 3.0. When the input image is supplied, only near to 35%(top half) of the image is colored gray(128,128,128). The rest of the image pixels still contain the original values. I tried various pictures of various resolutions, all yield the same result.

Only the top part of the image has been edited. The rest is still the original image.Resolution:1280x768 A sample output image

Answer 1

The behavior you're seeing is because you're not taking into account that BGR images have 3 channels. So for each pixel, there are actually 3 bytes that you would need to set to 128. One way to correct it would be to modify your inner loop to the following:

for (j = 0; j < (nCol * input_image.channels()); ++j)

Hope that helps.

Answer 2

Given only the top 35% or so gets changed, I'd say there's something wrong with your for loops which means you don't access the latter 65%. (Your image having multiple channels isn't the issue here, as changing only one of the channels would simply affect the colour balance, not only colour a certain portion of the image). However, there's really no need to recolour the image row by row.

Miki's suggestion in the comments of using setTo to change the values is good. Here's the documentation for more information. Repeating the code snippet anyway:

input_image.setTo(Scalar(128,128,128))

If you don't want to lose the original image, because loading an image just to make it grey does seem kinda strange, make a new grey matrix of the same size, using something similar to the following (John Wiggles did have a point):

Mat grey(input_image.size(), CV_8UC3, Scalar(128,128,128));

See the documentation for more ways to declare Mat types.

Answer 3

Wait... you want to simply fill/replace all values with 128? Why use the original image at all? If I am understanding you correctly, then simply do this:

1.) Create an empty Mat image with same size as the image.

      (Mat Mat_object = Mat(image.rows,image.cols,IMAGE_TYPE) )

2.) Initialize all values of the Mat object to 128. (Google is your friend)

3.) Create two more copies of that image (if you want multiple layers of such a structure)

               (  Mat_object.clone() helps here  )

4.) Create a:

         vector<Mat> object, then use object.push_back(image/copy1/copy2)

5.) Use merge(object,final_image)

6.) Enjoy your blank grey image

7.) Print your blank grey image

8.) Frame your blank grey image

9.) It is a blank grey image

wow