Abstract:

The existence of speckle noise has seriously affected the availability of synthetic aperture radar (SAR) images and poses serious challenges to the subsequent image segmentation, feature extraction and target recognition. Combining the features of nonsubsampled directional filter and dual-tree complex wavelet transform, a SAR image de-noise algorithm based on the advantages of the two algorithms mentioned above is proposed, which is a local hybrid filter under nonsubsampled directional filter bank and dual-tree complex wavelet transform de-noise algorithm. The new algorithm holds the features of multi-direction and multi-scale nature of signal, can keep the translation invariance of the image, and improves the image visual effects. The algorithm uses the dual-tree complex wavelet filter following the nonsubsampled direction filter. The difference is that the new algorithm can not only de-noise for the high frequency, but also for the low-frequency part of the multi-resolution multi-frequency band filter for the noise existed in both the high-frequency and the low-frequency. The experiment results show that, compared with the dual-tree complex wavelet-Contourlet transform and soft threshold method, the peak signal-to-noise ratio (PSNR) is increased by 2 dB, and compared with the Contourlet with cycle spinning and soft threshold method, the PSNR is better and the de-noise image is more smooth; moreover, man-made textures are restrained and a significant improvement in the visual effects is obtained.