Denoising of colour images using window contrast enhancement and vector alignment

In this paper a two stage filtering scheme for removal of random valued impulse noise from colour images is presented. The first stage consisting of an impulse detector employs iterative contrast enhancement in the filtering window. The performance of the impulse detector is further improved by exploiting the inherent correlation among three colour components. The second stage restores the noisy pixel by optimally aligning the pixel vector such that the brightness and the colour of the restored pixel are close to the original pixel. As the proposed method also includes the noise free component in the process of restoration of noisy components, unlike other component-wise filtering approaches where restoration of noisy components is done independently, colour artefacts in the present scheme are avoided. A significantly improved performance of the proposed detection and filtering approach is demonstrated through several experiments.