Abstract
A spacial image filter sequence is often used as a preprocess part in an image processing application and is important for enhancing features or removing unnecessary information from images. Designing such a sequence is a time-consuming work even if it is done by experienced engineers, therefore automatic generation has been studied for some decades. However, the robustness of the generated sequence and coexisting of image operations and numerical calculations in a sequence are still challenges. We propose a novel method based on a Multi-Objective Genetic Programming to generate a robust image sequence consisting of filters which processes Image-Value pairs by giving sets of images before and after processing. We demonstrate our method by generating a background elimination filter from flower pictures.
