Detecting spreaded white spots of light on a noisy gray-scale image.

edit

If your dot patterns are perfectly aligned (to within one pixel of a perfect grid), you can use cv::reduce to compute a row projection profile and a column projection profile respectively. Then, you can detect the peaks on each, and use those row and column indices to find the approximate position of each dot.

If your dot patterns are not perfectly aligned, there are other techniques applicable, with increasing order of difficulty of implementation.

You could do,

1. Threshold the image.

2. Find contour.

3. Calculate central point for each contour using Image Moments.