Speckle-noise reduction via "rotated elliptical-thresholding" in an homomorphic complex-wavelet domain

Speckle noise has been found to rotate an image´s homomorphic elliptically distributed complex-wavelet coefficients. This work proposes a counter-rotation of these complex-wavelet coefficients as a means of mitigating speckle noise using principle component analysis. Simulations suggest that the proposed denoising technique offers superior visual quality, though its signal-to-mean-square-error ratio (S/MSE) is numerically comparable to the improvement obtained in customary median filtering and adaptive Wiener filtering.