Image Denoising Using Adaptive Threshold Based on Second Generation Bandelets Transform

Second generation bandelets transform is a new method for image asymptotically optimal approximation based on capturing the geometric content of image. Relative to the first generation bandelets, the second generation is an construction of absolute discretization which can complete the image dissection through multi-scale analysis and geometric direction analysis. Bayes is an ideal estimation method of threshold. This paper perposes a novel denosing method based on second generation bandelets and bayes adaptive threshold, and makes full use of intrinsic geometry characters of image. Experiment on image denoising which the Lena and Barbara image are spliced with different level of gauss white noise shows that: compare to adaptive wavelet method, the algorithm perposed in this paper improves the peak signal-to-nosie ratio.