Abstract: The multiplicative speckle noise is widely found in sonar images, which seriously affects the image quality. The noise is characterized by rayleigh distribution. Aiming at this kind of noise, in combination of a total variation algorithm and the design ideas of the denoising algorithm of three dimensional block matching image (BM3D), a new denoising algorithm of total variation block matching sonar images is proposed. First, the algorithm uses 2-norm to complete the block matching group of noise sonar images. Then, the total variation algorithm is used for filtering and denoising of image block after grouping, due to the characteristics of fuzzy sonar image, less information of texture details, etc. Finally, the image after denoising is obtained by means of weighted clustering of the filtered image block. The experimental results show the effectiveness of denoising about the proposed algorithm in this paper. Moreover, it has better denoising effect compared with the classical Lee filter, Frost filter, BM3D and total variation algorithm.

Key words: sonar image, multiplicative speckle noise, total variation, image denoising, block matching