Abstract:

Combined with the characteristics of the coefficients in the Shearlet domain, an synthetic aperture radar (SAR) image denoising method is presented based on the sparse representation of coefficients and projected total variation (TV) method. The problem that the edge details of the image often lost in the processing of sparse representation and the staircasing effects caused by total variation can be resolved by the proposed method. Firstly, the sparse representation model of the SAR image is constructed and the stagewise orthogonal matching pursuit (StOMP) is used to obtain the optimization solution, which is the unbiased estimation of the real image’s coefficients in terms of the statistical mean. Secondly, to make up the loss of the image details result from the coefficients dropped in the sparse representation processing, the projected total variation scheme is given to iterative denoising, that utilizes the property of the dropped coefficients have the ability to characterize edges of Shearlet coefficients and projects the image to the Shearlet functions that corresponding to these dropped coefficients to get the reconstructed image. The experimental results demonstrate that the proposed method combines the sparse denoising and projection TV based on the characteristics of coefficients in Shearlet domain, that corrects their respective defaults not only suppresses the speckle but also achieves better performance in terms of effectively maintaining the image texture details and subjective visual quality.