Adaptive color image filtering based on center-weighted vector directional filters (Q1430204)

Color carries an essential part of information in many image recognition applications. Therefore, the improvement of colored images by noise filtering is an essential processing step. The corresponding methods can be subdivided into componentwise processing which operates the color channels separately, and vector processing which is more powerful because it considers the correlation between the color channels. This paper contributes to vector filtering methods. At first, the concept of vector directional filters is introduced. It is shown that different distance measures can be applied (angular distance measures and generalized Minkowski distances, resp.) which can be combined to form a hybrid vector filter also. Then, a set of weights is introduced when the partial distances are summed up which allows an improved adaptation to a specific problem. Basing on this co-called CWVDF approach (center-weighted vector directional filters), two improvements (adaptive CWVDF and switched CWVDF) are proposed and investigated. Extensive experiments with standard test images show that the proposed methods prove to be flexible and powerful in filtering.