۱٫ Introduction

۲٫ Image Super-Resolution Reconstruction Method Based On Sparse Residual Dictionary

۳٫ Experiment Analysis

۴٫ Conclusion

۵٫ Acknowledgment

۶٫ References

In order to improve the resolution of degraded images, an image super-resolution reconstruction method based on sparse residual dictionary is proposed. Firstly, the fuzzy matrix and down-sampling are used to degrade the high-resolution image set to get the corresponding low-resolution image set, and the Bicubic interpolation method is used to reconstruct the low-resolution image, and the high-resolution residual image set is obtained by comparison. The residual image contains the high frequency information of the image. Secondly, using the method of sparse dictionary learning, the residual maps are trained as a sample, and the sparse residual dictionary pair through joint dictionary training. Finally, the sparse coefficient of the image calculated by using the low-resolution dictionary and the low-resolution image to be reconstructed, and the similarity between the low-resolution and high-resolution image blocks and the sparse representation of the corresponding real dictionary strengthened, so as to realize the image super-resolution reconstruction. The experimental results show that the proposed algorithm performs well in both subjective and objective evaluation of reconstructed images.

Introduction

Image super-resolution (SR) reconstruction uses a single or a set of low-resolution (LR) to reconstruct a high-resolution (HR) image based on certain assumptions or prior information [1]. Since the input of low-resolution images provides limited information, image super-resolution reconstruction is a typical ill-conditioned inverse problem that requires relevant prior information [2]. Existing super-resolution methods can be roughly divided into three categories: based on interpolation [3], based on reconstruction [4] and based on learning [5]. The interpolation method based on polynomial approximation with full representation, the method is simple and fast, but the image blur quality is poor. The reconstruction-based method can effectively maintain the boundary sharpness and suppress false by applying a set of linear constraints to unknown high-resolution pixel values.