如何在OpenCV中遮罩浮点图像？

Question

I have a Nx1 matrix of floating-point values. I wish to get the average of only those values starting from the 22nd to the Nth item. I have created a binary mask of Nx1 dimensions, put 0s in the range [0,21] and 1s in the rest. Applying that mask while calculating the average (using inbuilt functions for maximum speed) gives an error. The code snippet is given below.

The error is:

sizes on input arguments do not match in cvSvg()"

What should be the correct mask? I have tried using floating-points instead of integers in the mask, still not working.

CvMat mask;
int i;
int N = img->width;
IplImage* W = cvCreateImage( cvSize(N, 1), IPL_DEPTH_32F , 1 );
IplImage* A = cvCreateImage( cvGetSize(img), IPL_DEPTH_32F , 1 );
cvConvertScale(img, A);
int* vals = (int*)malloc(N*sizeof(int));

for(i=0; i<N; i++)
{
    if(i<22)
        vals[i] = 0;
    else
        vals[i] = 1;
}
cvInitMatHeader(&mask, N, 1, CV_8U, vals);

cvSVD(A, W, NULL, NULL, CV_SVD_MODIFY_A);

CvScalar mean = cvAvg(W, &mask);

Answer 1

Here is how I computed the mask with a contrived example, but it appears to work just fine.

int main(int argc, char* argv[])
{
    int N = 50;
    CvMat* a = cvCreateMat(1, N, CV_32FC1);
    CvMat* mask = cvCreateMat(1, N, CV_8UC1);

    float* aData = a->data.fl;
    uchar* maskData = mask->data.ptr;

    for(int i = 0; i < a->cols; i++)
    {
        aData[i] = (float)i;

        if(i < 22)
        {
            maskData[i] = 0;
        }
        else
        {
            maskData[i] = 1;
        }
    }

    CvScalar avg = cvAvg(a);
    cout << "Average without mask: " << avg.val[0] << endl;

    avg = cvAvg(a, mask);
    cout << "Average with mask: " << avg.val[0] << endl;

    cvReleaseMat(&a);
    cvReleaseMat(&mask);
    return 0;
}

This produces the following output:

Average without mask: 24.5
Average with mask: 35.5

So, it is working for this simple example. Hopefully that will get you going again.

Answer 2

the image you use as mask does not have to be double, but Uint, the code is fine I think, not really sure about the initialization of the mask. still you must declare W as follow

IplImage* W = cvCreateImage( cvSize(N, 1), IPL_DEPTH_8U, 1 );