Summary
A data-adaptive fuzzy filtering framework is designed to remove noise in microarray images without the requirement for fuzzy rules and local statistics estimation, or under unrealistic assumptions that the original signal is available. This is achieved by utilizing the inference engine in the form of transformed distance metrics between the samples within the supporting window. The training of the filter coefficients is thus based on local image features. Proposed fuzzy filters can preserve important structural elements and eliminate degradations introduced during microarray image formation.