One of the most investigated problems which exist in image processing is to find the optimal method to extract the different features(edges, regions,...) of the image.
The main source of these problems is the diversity of images because they are obtained under different conditions of illumination and by different sensors. So a lot of algorithms are proposed to extract these features and any one of the algorithms gives good results in all cases with all images. So the idea is to find an algorithm which depends only on the used camera as the OTF (Optical transfer function). The majority of methods are based on the relation :
Image output = Image input * filter
So
a 2 Image output = Image input * a 2 filter axay Nay
In practice we use directly the second relation, we apply a convolution product between the image and a mask, in reality the coefficients of the mask are the derivative coefficients of the filter. This relation shows that the image is always forced to be filtered with a certain filter as gauss filter, mean filter, ...etc . In reality the final contour image is not of the original image but it is the contour of the filtered image. So my method is different, I use a relation which is obtained with image restoration model including the OTF function.