Adaptive block threshold and tight frames for image denoising
Yibin Yu, Xiao-Ben Zheng, Junying Gan · 2011
In this paper, we propose an Adaptive BlockThresh-old and Tight Frames Denoising (ABTTFD) using sub-block threshold for image denoising in wavelet do-main. The size of sub-block is equivalent to LLn where LL is the low frequency subband, n is the level of wavelet transform. The proposed thresholding called Block-Shrink is more efficient and adaptive because the threshold depends on wavelet coefficients of sub-block and uses the scale parameter. Moreover, tight frames are used to eliminate artifacts. Several test images are used for comparison among several denoising techniques like ABTTFD, Soft thresholding (Soft) and Block translation invariant stein thresholding (B-stein). Experimental results show that ABTTFD not only has good performance in visual effect but also obtains higher SNR and computing efficiency. Virtually, it outperforms Soft and B-stein.