Empirical study on sparse representation model of SAR images
An example of using the sparse representation algorithm to obtain a basis function dictionary for synthetic-aperture radar(SAR)scenes.The comparative of sparse representation of images before and after filtering shows that speckle noise has an impact on the dictionary results of sparse representation of SAR scenes.Select specific SAR image data from Pujiang No.2,ALOS2,and SIR-C,it is discussed that the effects of optimization algorithm,sample content,dataset size,radar resolution,polarization method,and band on dictionary results by setting single factor conditions.The results show that:(1)The dictionary learned from sparse representation of SAR scenes is related to radar band,resolution and polarization mode,and is independent of sample contents,datasets size,and optimization algorithms.(2)C-band can better reflect the sparsity of SAR scenes than L band.(3)The downsampling dataset can better reflect the sparsity of SAR scenes.(4)The dictionaries learned from HH and VV polarized images have more essential features.
SARsparse representationbase function dictionaryspeckle noisepolarization
国家科技期刊平台
NSTL
万方数据
