Sparse representation SAR image despeckling method based on Gaussian scale mixture model

Abstract

The invention relates to a sparse representation SAR image despeckling method based on a Gaussian scale mixture model, and belongs to the field of image filtering research. According to the method, firstly a sparse representation model of a single image block is established; then the sparse coefficient alpha is represented by the GSM model according to the statistical property of coherent speckles and the Bayesian estimation principle so as to obtain an optimized model; meanwhile, SAR images are classified and a sparse model is established according to the classification result; and finally the model is solved by using a convex optimization method so that the optimal sparse representation can be obtained and thus the denoised images can be obtained. According to the method, the coherent speckles can be effectively suppressed and the detailed information of the SAR images can be greatly reserved so as to achieve the great image reconstruction quality.

Description

technical field [0001] The invention relates to a sparse representation SAR image speckle reduction method based on a Gaussian proportional mixture model, which belongs to the field of image filtering research. Background technique [0002] Synthetic Aperture Radar (SAR) has become a representative radar system in the field of microwave remote sensing due to its high-resolution imaging capabilities in azimuth and range, and has attracted widespread attention in both military and civilian applications. When the radar emits electromagnetic signals to irradiate the target, the interference between the random scattering signal of the target and the transmitted signal will produce coherent speckle noise, which seriously affects the image quality, so the suppression of coherent speckle noise is particularly important. The most direct way to eliminate speckle noise is to use multi-view processing, but this method will reduce the azimuth resolution. A more reasonable way is to use f...

AI Technical Summary

Problems solved by technology

However, although the PPB algorithm can suppress the coherent speckle better, it will introduce a large number of artificial artifacts, resulting in a large loss of image detail information. The SAR-BM3D and FANS algorithms are relatively better in maintaining detail information, but the speckle reduction effect is not as good as PPB. , unable to suppress coherent speckle noise as much as possible

Images