Method and device for image stylized reconstruction

Abstract

The present invention provides an image stylized reconstruction method and device. The method comprises the steps of obtaining multiple pairs of source style images and target style images of same scenes; according to the multiple pairs of source style images and target style images of same scenes, obtaining a source training set Y and a target training set X; training the source training set Y and the target training set X separately, determining a source style dictionary Dy, a target style dictionary Dx and a mapping relationship between a sparse coefficient alpha y corresponding to the source training set Y and a sparse coefficient alpha x corresponding to the target training set X, reconstructing a to-be-processed source style image y' by a sparse representation method according to the source style dictionary Dy, the target style dictionary Dx and the mapping relationship M, and obtaining a target style image x' corresponding to the to-be-processed source style image y'. According to the present invention, the reconstruction from the source style image of any style to the target style image of any style is realized, and the reconstruction accuracy is high.

Description

technical field [0001] Embodiments of the present invention relate to the field of image processing, and in particular, to a method and device for image stylized reconstruction. Background technique [0002] In real-life applications, there are many different imaging modes and corresponding style images for different purposes, such as near-infrared images used for face recognition with constant illumination; sketch images used in law enforcement or entertainment equipment ; Thermal infrared images used to detect life in special situations such as earthquakes, etc. [0003] However, in order to meet the requirements of some image applications, it is necessary to convert images of different styles in the same scene into images of the same style. For example, in face recognition, the matching effect between face images of different styles is very poor. After converting face images of different styles into face images of the same style, the accuracy of face recognition can be i...

AI Technical Summary

Problems solved by technology

[0004] The existing image stylized reconstruction technology is mainly based on the brush stroke image stylized reconstruction method. This image stylized reconstruction method can only reconstruct an image of a specific style into an image of another specific style. The generality of image reconstruction is relatively low Poor, and the outline of the reconstructed style image is blurred, and the accuracy of image stylized reconstruction is poor

Images