How to load nearby pixels using pointer - c++

Question

Suppose i have an image matrix and i am at a particular pixel [say 4] like this:
0 1 2
3 `4` 5
6 7 8
I am trying to cycle through all pixels and am attempting to access 0,1,2, 3,5 6,7,8 whose values i am storing in the array called Pixel.... here is my attempt at it using OpenCV, kindly tell me where am i going wrong.
I am using pointer temp_ptr to access the IplImage image.
uchar* temp_ptr=0 ;
CvScalar Pixel[3][3];
int rows=image->height,cols=image->width,row,col;
for( row = 0; row < rows-2; ++row)
{
for ( col = 0; col < cols-2; ++col)
{
temp_ptr = &((uchar*)(image->imageData + (image->widthStep*row)))[col*3];
for (int krow = -1 ; krow <= 1; krow++)
{
for (int kcol = -1; kcol <= 1; kcol++)
{
temp_ptr = &((uchar*)(image->imageData + (image->widthStep*row+krow)))[(col+kcol)*3];
for(int i=0; i < 3; i++)
{
for(int j=0; j < 3; j++)
{
for(int k=0; k < 3; k++)
{
Pixel[i][j].val[k]=temp_ptr[k];
}
}
}
}
}
}
}
I am not really sure how to load the sorrounding Pixels usingtemp_ptr, please help me out.

Answer 1

Well sir, it sounds like you want to do convolution, and doing it this way when you have OpenCV at your fingertips is a bit like hammering a can opener on your Spaghettios to burst it open by blunt force.
In fact, what you're doing is almost exactly the output of cv::blur(src, dst, cv::Size(3,3)) except it also includes the center pixel in the average.
If you want to exclude the center pixel then you can create a custom kernel - just a matrix with appropriate weights:
[.125 .125 .125
.125 0 .125
.125 .125 .125 ]
and apply this to the image with cv::filter2d(src, dst, -1, kernel).

Answer 2

Assuming image->imageData is in RGB format, so there are 3 bytes for each pixel, you could do something like this:
int rows = image->height;
int cols = image->width;
uchar* temp_ptr = 0;
CvScalar pixels[8];
for (int col = 0; col < image->height - 2; col++) {
temp_ptr = image->imageData + image->width * col + 1;
for (int row = 0; row < image->width - 2; row++) {
temp_ptr += row * 3;
pixels[0].val = temp_ptr - width * 3 - 3; // pixel 0 from your example
pixels[1].val = temp_ptr - width * 3; // 1
pixels[2].val = temp_ptr - width * 3 + 3; // 2
pixels[3].val = temp_ptr - 3; // 4
pixels[4].val = temp_ptr + 3; // etc...
pixels[5].val = temp_ptr + width * 3 - 3;
pixels[6].val = temp_ptr + width * 3;
pixels[7].val = temp_ptr + width * 3 + 3;
// calculate averages here and store them somewhere (in a vector perhaps)
}
}
Note I didn't test this code.

Answer 3

First of all you have to start learning some programming. Your complete code is a mess.
Some major problems I could quickly found:
First of all you have to start your first two for loops from 1 (because you decrement by -1 when you apply the window) and you will end up reading some memory address that are not allocated.
Second the first temp_ptr = &((uchar*)(image->imageData + (image->widthStep*row)))[col*3] is useless so you can remove it.
the other
temp_ptr = &((uchar*)(image->imageData + (image->widthStep*row+krow)))[(col+kcol)*3];
is having a small problem, the operator precedence, should be:
temp_ptr = &((uchar*)(image->imageData + image->widthStep*(row+krow))[(col+kcol)*3];
you don't need the other 3 inside loops
Also is not clear what you want to do, you want to get the neighborhood of a specific pixel (then you need no loops) or you want to apply a kernel to each pixel from the image.