Super-resolution image reconstruction method based on codec structure

• Main Authors: KONG MENGXING, ZHANG JING, SHAO MINHAO
• Format: Patent
• Language: Chinese; English
• Published: 03.12.2019
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention provides a super-resolution image reconstruction method based on a codec structure. The super-resolution image reconstruction method is used for solving the technical problems that in the prior art, the peak signal-to-noise ratio and the structural similarity of a reconstructed high-resolution image are low. The method comprises the following steps: constructing a codec network; obtaining a training sample set composed of high-resolution and low-resolution image block pairs; training the coder and decoder; and performing super-resolution image reconstruction on the RGB image to be reconstructed by using the codec network. In the multi-scale feature fusion network, the downsampling modules which are sequentially connected in a stacked mode are adopted to extract the multi-scale features of the low-resolution image, and the upsampling reconstruction module is used for nonlinear mapping, so that the peak signal-to-noise ratio and the structural similarity of the reconstructed high-resolution image