An Entropy Minimization Approach for Designing W-Operators

The task of solving an image processing problem is often very complex. It depends on prior knowledge of the problem domain and on the knowledge, experience and intuition of an image processing specialist. Motivated by this complexity, some research groups have been working on the development of automatic techniques which try to mimic the image processing expert. Previous works in this area include [1] and [2]. The goal of these techniques is to have a system which receives a collection of image pairs that represent the desired input and output of the image processing operation, and, based on these example images, designs an image operator with behavior similar to the illustrated in the examples. The existence of such a system would also make the design process accessible to professionals who are not familiar with the underlying technical details of the image processing field.