Comparison of intensity image denoising methods based on non-local filtering algorithms for real coherent ladar

The denoising algorithms based on non-local filtering can obtain higher quality of restored images than the classical ones based on local statistic characteristics. Due to the influence of speckle noise on intensity images of coherent ladar, the non-local filtering algorithms are integrated with homomorphism transform in this paper, i.e., homomorphism non-local means (H-NLM) algorithms, homomorphism multilevel non-local means (HM-NLM) algorithms, homomorphism non-local means by blocks (HB-NLM) algorithms, and homomorphism block-matching and three-dimensional filtering (H-BM3D). Real intensity images of coherent ladar are denoised by these algorithms, and the results are compared and analyzed. It is shown show that single frame intensity image denoised by H-BM3D and 6-frame averaging intensity image denoised by HM-NLM can achieve better restoration. The restored images have better gray uniformity in target area and better clarity of edges, and can reflect the geometry structure of target.