Accessing IplImage element of type IPL_DEPTH_16S in OpenCV

I know that for a single channel byte image you do:

((uchar *)(img->imageData + i*img->widthStep))[j]

and for a single channel float image, you do:

((float*)(img->imageData + i*img->widthStep))[j]

But how about for 16 b开发者_StackOverflow社区it signed images (IPL_DEPTH_16S), I tried:

((short*)(img->imageData + i*img->widthStep))[j]

and

((signed int*)(img->imageData + i*img->widthStep))[j]

to no avail.

Thanks,

---

Actually, the short works fine if you take widthStep/2...

---

The answer is :

 (((short*)(img->imageData)) + i*img->widthStep)[j]

And the reason is explained by this example:

#include <stdio.h>

int main(){
    char * pointer;

    printf("%zu \n", sizeof(char));
    printf("%zu \n", sizeof(signed short));
    printf("%zu \n", sizeof(signed int));
    printf("%zu \n", sizeof(float));

    printf("%p \n",((char*)(pointer) + 10 * 5));
    printf("%p \n",((signed short*)(pointer)) + 10 * 5);
    printf("%p\n",(((signed int*)(pointer)) + 10 * 5));
    printf("%p\n",((float*)(pointer)) + 10 * 5);
}

1 
2 
4 
4 
0x7fff5fc01084 
0x7fff5fc010b6 
0x7fff5fc0111a
0x7fff5fc0111a