A sharpening vector median (VM) filter that simultaneously denoises and enhances vector-valued image signals is introduced in this paper. This filter makes use of the trimmed aggregated distance minimization concept and robust vector order statistics as a way of enhancing edges and image details while retaining the effective noise removal characteristics of the standard VM operator. Depending on local image statistics and the user’s needs, the new procedure accommodates various design, implementation, and application objectives by enhancing the vector-valued signals. The filter properties claimed in this paper are carefully demonstrated and the performance and efficiency of the new filter are critically analyzed. The new filter exhibits essential filtering properties, including root signals, zero impulse response, and invariance to rotation, scale, and bias. The sharpening vector median filter appears to be a robust operator. Examples of color image filtering and virtual restoration of artworks are included.
The paper is carefully written, thorough, and well organized. The primary sections are devoted to the sharpening issue, the rationale for the proposed approach, and extensive performance analysis. There are tables and figures throughout to document the approach and results. A shorter section addresses the issue of computational complexity for the filter. As should be expected, the technical details presented are notationally complex, but logically and carefully handled. The paper should be of interest to anyone interested in improvements to VM filters for image processing.