Image denoising algorithm based on DTCWT and adaptive windows

In this paper, a new image denoising algorithm based on dual-tree complex wavelet transform (DTCWT) is proposed, in which directional windows are chosen as local neighborhood to estimate the variance. For the different subbands within the same scale, better estimation about energy cluster can obtained by ellipse windows than square windows, and the sizes of the ellipse windows for different scales are also different. The experimental results indicate that the method of choosing directional ellipse windows is uncomplicated and effective. Compared with the LAWMAP method, PSNR gained by our proposed method is higher by 0.7 dB at most, and visual quality is also improved.