The research on image super-resolution plays a significant role in enhancing image quality and improving the utilization of massive existing image resources. The application of convolutional neural networks to super-resolution models has greatly improved the quality of super-resolution image, and the introduction of residual networks has made it possible to train more complex models. However, speeding up the processing while ensuring the quality is still a problem that needs to be solved. Based on convolutional neural network and residual structure, we propose a super-resolution model with multi-feature extraction module, considering both image quality and algorithm speed. Simultaneously extracting structural features and detail features of images, we increased processing speed more than 40% while ensuring the quality.
Paper
Full text
Fast Image Super-Resolution Based on Multi-Feature Extraction Network
Semantic Scholar · Computer Science · 2022
Abstract
The research on image super-resolution plays a significant role in enhancing image quality and improving the utilization of massive existing image resources. The application of convolutional neural networks to super-resolution models has greatly improved the quality of super-resolution image, and the introduction of residual networks has made it possible to train more complex models. However, speeding up the processing while ensuring the quality is still a problem that needs to be solved. Based on convolutional neural network and residual structure, we propose a super-resolution model with multi-feature extraction module, considering both image quality and algorithm speed. Simultaneously extracting structural features and detail features of images, we increased processing speed more than 40% while ensuring the quality.