Image denoising based on sub-image restoration through threshold optimisation
Somnath Mukhopadhyay, Jyotsna Kumar Mandal · International Journal of Computational Intelligence Studies · 2015
This paper proposed a novel approach for detection and filtering of impulses in digital images through median filtering based on optimisation through particle swarm optimisation (PSO) technique. The input image is logically partitioned into 5 × 5 sub images. Two restored versions are generated through proposed technique using 3 × 3 and 5 × 5 masks separately of which better one in terms of fitness is chosen. The detection method is based on all neighbour directional pixels. The filtering method is based on weighted median approach. Out of the three user parameters used for detection and filtering, a single one is searched in a wide range using PSO to obtain the optimal solutions. The results obtained demonstrate the effectiveness of the proposed technique. On both the noise models same type of results and comparisons have been carried out and time requirement for the techniques have also been studied and compared.