A simple real-time image cloning algorithm based on modified mean-value coordinates

Image cloning which could copy the source image and “paste“ it into the target image seamlessly is one of an important digital image manipulation technique. Traditional methods to cope with this problem were to analyze image information in gradient domain according to solving Poisson equations. As a result, existing methods need high computational cost and memory usage which hinder them from real-time applications with high-resolution images. In this paper, we propose an algorithm to cope with image cloning problem based on modified mean-value coordinates technique so that we transform the problem into an image interpolation procedure which is very simple and easy to implement without diminishing the performances. Furthermore, we use GPGPU (General Purpose Computing on Graphics Processing Unit) technique to accelerate the proposed algorithm and achieve real-time performances with the cloned region whose size is about 1 mega pixels.