Image de-noising method based on HMT model interlayer mapping in wavelet domain

A new image de-noising algorithm based on image segmentation is proposed to keep image edges more effectively. The proposed method segments the low-frequency subband into many domains adaptively by PCNN, and treats the connected regions gotten as neighbourhood. With a simplified HMT model in both discrete and stationary wavelet, the quad-tree inter-layer model is used to map the neighbourhood into the highfrequency subbands. And the regions gotten are taken as irregular neighborhoods for denoising. Further it chooses coefficients both in irregular neighborhood and in a fixed rectangular window, and selects the coefficients have closer geometric distance. A better restoration of images is demonstrated in the results of experiments, with detail of images kept as well as image noises decreasing.