Just a general question regarding contour detection. I have found several examples where dilation and erosion is made sometimes before canny or sobel and sometime after. Which is the correct approach ? In my scenario I have 2048x2048 images with very little objet to detect, very close each other.