Impulsive noise filtering based on noise detection in corrupted digital color images (Q5957410)

The paper describes new approaches for removing corrupted pixels in digital color images. This process is performed in two steps. At first, the algorithm identifies corrupted pixels. This step avoids interfering with the non-corrupted pixels, which results in better quality and computational efficiency. Secondly, two new detection-estimation based algorithms are introduced which are applied to the classified pixels. The first algorithm is a scalar detection approach, and the second is a vector-based one. The thresholds which are used for the comparison follow Weber's law for the sensitivity of the human eyes. The paper includes experimental results as well as comparison tests with other methods which show that the proposed methods combine comparable or better quality with high performance.