Threshold pattern filters for image enhancement
Pak-Cheung Lai, Bing Zeng · 1999
Threshold Boolean filters (TBFs) constitute large class of non-linear filters that are very effective in removing impulsive noise. However, the required computational complexity is also high because thresholding has to be done at all levels. In this paper, a new kind of filters based on threshold operation is proposed. The proposed filters, referred as threshold pattern filters (TPFs), threshold the filter window only at some levels. A TPF picks a sample within the window as the output based on the threshold patterns obtained. An efficient MAE-based training method for designing TPFs is proposed. Simulation results show that the performance of TPFs is comparable to that of the corresponding TBFs, when they are applied to enhance images corrupted by mixed noise.