Speckle Noise Removal by Total Variation and Curvelet Coefficient Shrinkage of Residual Noise
Susant Kumar Panigrahi · 2019
This paper proposed a Spatio-Frequency approach to remove multiplicative noise using total variation (TV) and Curvelet coefficient shrinkage of the residual noise. Prior to that, it also experimentally validated the importance of phase in speckle noise removal. We have shown that the phase of any complex transformation is more immune to speckle, than its magnitude. The proposed analysis on different phases encouraged the authors to apply threshold on Curvelet magnitudes to obtain the suppressed image details due TV minimization. The experimental results of our algorithm (on the simulated speckle noise) are compared with some of the recently reported image despeckling techniques. Two standard measures namely: Peak Signal to Noise Ratio (PSNR) and Structural Similarity Measure Index (SSIM) are adopted to analyze these methods quantitatively. Finally, the reference (noise-free) images of 3 different databases (with more than 500 images of wide variety of natural scenes, human faces, objects and textures) are considered to generate noisy image with varying number of looks; L= 4, 8, 12,16. The obtained denoising results demonstrate that the proposed technique provides significant improvement in noise suppression with high PSNR and SSIM values compared to other methods.