Design of an efficient space transform domain based denoising technique for grey scale images

Abstract

In the present work, our primary objective is to design effective denoising algorithms in the space-transform domain to substantiate the need for consistency in denoising performance over wide range of noise levels with removal of residual noise. For this, we propose to develop image denoising algorithm based on bit plane slicing and bitonic filtering. The fundamental conjecture of segregation of noise in eight bit-planes and adaptive filtering is able to depict consistent performance at varying noise levels. Further we have optimized the performance of the proposed algorithm in order to demonstrate increased PSNR values with lower level of complexity. A novel method for estimation and removal of residual noise has also been proposed for optimum feature recovery. With the help of these hybrid domain methods we are able to depict consistency in denoising performance, preserve fine feature details and remove residual noise. With comparative analysis, it has been established that proposed denoising methods surpass the performance of recent state-of-the-art denoising methods in terms of qualitative and quantitative evaluation both over wide range of noise levels.