Development of Improved Image Cloning Technique

Abstract

There are certain changes in world which fall below the visible spectrum and are difficult to be perceived through naked eye. These subtle changes are most of the times considered for useful purpose and also contain vital information. Schemes being used throughout are necessary to be applied in order to implement these changes; image editing is one of them and has been used widely. Our goal is to devise a technique that blindly set one image or its part into another image providing/carrying meaningful information. Due to the rapid advancement in technology as well as in image processing yet there is a need of efficient and reliable technique to make changes. In this thesis, Modified Poisson Blending (MPB) is used as baseline and same is further modified to propose new one. The proposed scheme utilizes super-pixels segmentation technique to reduce the artifacts and hence to maintain color of source image in blended one that is reconstructed via Poisson editing. Further statistical parameters of patch are adjusted to retains its features. Image derivatives are performed to solve blending problem that is main task of this algorithm. Blended images are then evaluated using Natural Image Quality Evaluator (NIQE), Structure Similarity Index Measure (SSIM) and Peak Signal to Noise Ratio (PSNR) techniques and through visual evaluation to provide evidence that our proposed scheme outperform, . After evaluation, it is cleared that proposed technique perform better than pre-existing ones, since it avoids from artifacts that might appear at boundaries of blended images alongwith color information of source image.