A new technique for guided filter based image denoising using modified cuckoo search optimization

Abstract

In this work, a novel and efficient approach for image denoising is proposed. More often, noise affecting the pixels in image is Gaussian in nature and uniformly deters information pixels in image irrespective of their intensity values. This behaviour of noise can also be identified as Additive White Gaussian Noise (AWGN). For restoration of AWGN affected images, the proposed denoising approach is inspired by image adaptive guided image filtering using modified cuckoo search algorithm. The guidance image is itself derived from the noisy image for this purpose. Bilateral filtering smoothed noisy image is sharpened by unsharp masking and then employed as guidance image for the proposed optimal guided filtering approach. Optimal evaluation of parameters like guided filter smoothing parameter (regularization parameter or degree of smoothing (DoS)) and guided filter’s neighbourhood (kernel) size is done appropriately with the help of the modified cuckoo search algorithm. Two-dimensional search space is explored and exploited for deciding the behaviour of guided filtering adaptively as per the input image requirements. This guided image filter has a better behaviour at it acts as an edge preserving smoothing operator. It is considerably effective as its computational complexity is independent of filtering kernel size. A novel attempt is made by incorporating the Markov Random Field based Energy Minimization based objective/fitness function for imparting adaptive image denoising using metaheuristic intelligence. The proposed method is tested in terms of the performance metrics like peak signal to noise ratio, structural similarity index and mean square error. Performance of the proposed approach is compared with the already proposed image denoising techniques. For this comparison, only those methods are considered which were proposed for filtering of Gaussian Noise. Qualitative (visual) as well as quantitative (objective) results underlines the efficacy of the proposed method for filtering of Gaussian Noise.