Learning Based Adaptive Denoising Approach for Image Interpolation

In this paper, we propose an effective image interpolation framework through learning based adaptive denoisng approach. In the local area, error pattern between original image and interpolated image is treated as stationary Gaussian distribution. Under the initial estimation, the proposed method apply the patch as the basic unit, in which Multiclass SVM classifier is used to determine iteration number and denoise parameters. There are two steps in iterative processing, including adaptive denoise and data fusion. Experiment results shown the proposed method can significantly improve the interpolated image quality both subjectively and objectively.