Efficient Wavelet Based Denoising Technique Combined with Features of Cyclespinning and BM3D for Grayscale and Color Images

Denoising of images is considered to be a crucial preprocessing step in all image processing applications. Although various studies have been conducted in the area of denoising still there is room for improvement. A major challenge in denoising is tradeoff between noise removal and preserving edges. The proposed work attempts to achieve this by combining spatial and transform domain techniques. BM3D is the best denoising filter in spatial domain as it shows noticeable results when compared with other existing methods. Whereas wavelet based denoising preserves fine edge information. Biorthogonal wavelets are good for image denoising. Hence, this work is an attempt to combine these two methods. Firstly, check the performance of biorth 4.4 wavelets with biorth 6.8 by applying different thresholding techniques. It is observed that cyclespinning of images helps in achieving good denoising. Finally, bm3d and biorth 6.8 with cyclespinning feature is proposed which improves the performance of denoising for grayscale images. For color images, same procedure is done with cbm3d and biorth 6.8 in PCA colorspace. The results obtained are verified with quantitative and qualitative performance measures such as PSNR, MSE, and SSIM. Research results show that the proposed method is robust and efficient.