Image super-resolution reconstruction by fusing feature classification and independent dictionary training

Abstract

Super-resolution reconstruction plays an important role in reconstructing image detail and improving image visual effects. A new effective super-resolution method is proposed. Firstly, we extract the geometric features of the image patch to construct the decision tree, which will be used in patch classification in a supervised way. Then, we train the high-resolution and low-resolution dictionaries based on K-SVD independently for different types of training sets. Finally, we solve the mapping matrix for the coefficients between the high-resolution and low-resolution training set, which are used to map the low-resolution coefficients to high-resolution coefficients during the reconstruction phase to ensure accurate and fast reconstruction of the image patches. The experimental results show that the proposed method has a significant improvement in the reconstruction effect compared with other classic methods.