Abstract

Nonlinear image processing is presented here as a generalization of the operation by removing the linearity constraints. It seeks the optimum mapping implemented directly in logic. Given this generalization, the optimum nonlinear solution will be either better or equivalent to the linear solution, but it should not be worse. This inequality holds regardless of the problem or the criteria, provided that the training data is sufficient. The filter is an estimator it uses the input values to estimate an unobserved quantity. By making simple assumptions about the image statistics, we can estimate the output value at a specific point by considering only a finite window of observations centred at that point. Image processing is a very computing intensive task, because several low level (pixel level) operations are performed over an image in order to execute a certain task, like edge detection, edge linking, noise removal, dilation erosion and filtering. One common alternative is the use of dedicated DSP processors another is the use of an ASIC. Recently, another approach starts to be used the reconfigurable systems. However, such systems yet need some evolution to be used some development in the reprogram ability techniques and some improvement in their compilers.
