Image in Painting through Non Local Total Variation by Flower Pollination Approach and Predictive Guided Patch Mixing

In this paper, we improved the formulation of exemplar-based image inpainting using metric labeling by flower pollination optimization. In FPA, we used greedy approach for optimization of metric convergence in exemplar method, which increases the total variation, cost but reduce the convergence time. For reducing the cost, we used optimize number of masked images selected by Exception maximization method, which reduces the cost and increase the efficiency of total variation method. We used the parameter of quality score and cost on different type of four images and analyzed the PSNR, quality score in comparison with existing method. Experimental results show that the proposed approach significantly improves the MSE, PSNR and quality as compared to the existing method. There is 8-9% increase in quality score of inpainted images then proposed method and also 25-30 % increase in PSNR then the proposed method.Computional complexity is reduced in proposed method which in turn reduces time.