关键词: 图像去噪, 整体变分（TV）模型, 小波包分解, Bregman迭代

Abstract: The Total Variation（TV） model of Rudin et al. for image denoising is considered as one of the best denoising models. Theory suggests that the TV model denoises well piecewise constant images. For the texture rich image, in this paper, it is decomposed by multi-wavelet packet into a series of approximate piecewise constant sub-images. Sub-images are processed separately by TV model. Thus, the texture detail of image is preserved better and the Peak Signal to Noise Ratio（PSNR） of denoised image is improved 1 dB, compared with using TV model alone. At the same time, the improved Bregman iteration scheme is adopted to solve TV model. The algorithm convergence time has been greatly decreased.

Key words: image denoising, Total Variation（TV） model, wavelet packet decomposition, Bregman iteration