Image Interpolation Based On Edges Reconstruction

Image interpolation constructs large images based on smaller ones. If the interpolated images are clear in important details and edges, corresponding interpolation methods can be applied to reducing the transferred data in image transfer through network.We mainly introduce an efficient image interpolation method based on image edges reconstruction. The method firstly creates an auxiliary image from a gradient domain which is constructed by applying simple interpolation methods to the gradient domain of the input image. Secondly it applies bi-lateral filtering to the auxiliary image and extracted the edges from the filtered result by Canny edge detection algorithm. After then it compares the edges extracted with the edges of the input image to construct the final reconstructed edge set. In the last stage, it sets the image interpolated by simple methods (bi-linear interpolation for instance) as initial state and pixels close to the reconstructed edges and accurately interpolating the original image as static set. And the diffusion process is iterated over the initial state within a time limit to produce the output image.This method outperforms simple interpolation methods with regard to the performance around the edges. It’s also more computationally efficient compared with vectorization methods based on diffusion curves or subdivision curves. The method can easily be adapted for parallel computing which is preferred by applications on network like video chatting and interactive design...