Multiple Evolvable Hardware Image Filters by Analyzing Noise Types with Fuzzy Relations

Image filtering, which removes or reduces noises from the contaminated images, is an important task in image processing. In the recent years, evolutionary design of image filters that provides adaptive and hardware implement able solutions has been received a lot of attentions. This study deals with the design of multiple evolvable hardware (EHW) based image filters using fuzzy relations. Two indicators, similarity and divergence, are defined as fuzzy sets for describing the relations of pixels contained in a sliding window. In the proposed method, each pixel to be recovered is analyzed by the fuzzy relations and labeled as the associated noise type. Multiple EHW-based image filters, each of which is trained supervisedly by the pixels belonging to the same noise type, are built simultaneously. Because each image filter is dedicated to a specific type of noise, it can recover pixels of the noise type more accurately. With the proposed method, the efficiency of training EHW models and accuracy of image filtering are both improved. This paper evaluates and compares the performance of the proposed method with other ones. To our best knowledge, this is the first attempt to use fuzzy relations for noise categorization for the design of EHW-based image filters.