International Journal of Applied Mathematics and Machine LearningAuthor's: Shuaiqi Liu, Yu Zhang, Xiaole Ma, Ming Liu, Shaohai Hu and Jie Zhao
Pages: [135] - [145]
Received Date: March 12, 2016
Submitted by: Bin Guo.
DOI: http://dx.doi.org/10.18642/ijamml_7100121639
Speckle suppression is a hot research field in the world. In this paper, combining the advantages of shear filtering and dual tree complex wavelet transform (DTCWT), we propose a new SAR image de-noising algorithm based on K-singular value decomposition (K-SVD) and soft thresholding. The new algorithm uses shear filtering to do directional decomposition. After that, DTCWT is applied to do scale decomposition. Soft thresholding is applied to high frequency coefficients, while K-SVD is applied to low frequency coefficients. The new algorithm can not only remove most noise in the high frequency coefficients, but also remove a small amount of noise contained in the low frequency. The experimental results show that our method can both improve peak signal to noise ratio (PSNR) and enhance the image visual effect.
SAR de-noising, shearlet, DTCWT, K-SVD.
