SAR image denoising method based on sparse representation

Abstract

The coherent nature of radar illumination causes the speckle effect, which gives the synthetic aperture radar (SAR) image its noisy appearance. The probability distribution of speckle noise is multiplicative rather than additive, which makes the interpretation and processing of SAR imagery more difficult. A novel SAR image denoising method is proposed. First the multiplicative noise is transformed into additive-like noise by logarithmic transformation. After that, a novel object function is proposed which combines a pre-trained dictionary model to deal with the image. Finally, exponential transform is employed to recover the image. Experimental results show that the proposed method can effectively remove the noise of SAR images, and indicate good performance compared with other state-of-the-art methods.

Keywords

• image denoising
• synthetic aperture radar
• radar imaging
• speckle
• image representation
• statistical distributions
• image reconstruction
• transforms
• sparse representation
• coherent nature
• radar illumination
• speckle effect
• synthetic aperture radar image
• noisy appearance
• probability distribution
• speckle noise
• sar imagery
• multiplicative noise
• logarithmic transformation
• exponential transform
• sar image denoising method
• additive-like noise
• pretrained dictionary model
• image recovery