Super-resoloution image reconstruction based on wavelet packet transform and artifcial neural networks
Pann Ei San, Fengchun Tian, Zhiyong Shi, Minjun Deng · 2014
Super-resolution image reconstruction is an image processing technique that attempts to reconstruct high quality and high-resolution images from one or more low-resolution images by learning from a collection of training images. In this paper, new image resolution enhancement methods using wavelet packet transform and neural networks are proposed. The input image is decomposed by using wavelet packet transform. In this work, the wavelet packet decomposition sub-images are used to train neural networks. Multilayer Perceptron (MLP) and Radial Basis Function (RBF) neural networks are employed to predict the expected wavelet packet sub-images of a high-resolution image. The super-resolved image is finally produced by using the synthesis procedure of wavelet packet transform. The objective and subjective quality assessments indicate that the proposed methods outperform the conventional image resolution enhancement techniques.