Implementation of Image Inpainting using OpenCV and CUDA on CPU-GPU Environment

The procedure of filling lost data in digital images is an important perspective in image processing. Image inpainting refers to rebuilding techniques used to expel unwanted patches or restoration of an image. In this work, we are implemented the image inpainting using exemplar method on CPU and GPU environment. Exemplar-based method contains the fundamental procedure required to recreate both texture and structure. The strategy introduced here consolidates the qualities of both methodologies into a solitary, productive algorithm. The output is generated using exemplar method, is centered around a patch based filling way to deal with enhance execution speed and exactness of the geometric structures. In this work, we evaluated and compared the performance of exemplar method using parallel programming model OpenCV and CUDA on two platforms such as CPU and GPU. The experimental result shows that exemplar method on GPU based CUDA yields superior results when compared against the Quadcore Intel i3 and Hexacore Intel Xeon CPU processors for different patch size of deteriorated image. Since GPU has its own memory and a large number of cores, our result shows that implementation of exemplar method on GPU yields average improvement of execution against i3 over 90 times and Hexacore by 80 times approximately.