Residual Frequency Content Awareness Approach for Image Super Resolution

In the past few years, several advanced techniques for enhancing the resolution of single images, known as single-image super-resolution (SISR) network methods, have emerged. Nevertheless, the majority of current approaches fail to fully exploit the information that is available before and after the convolution process, as well as the high-frequency details present in the image. In this exposition, we introduce a concise and precise super-resolution algorithm, the Residual Frequency Con-tent Awareness (RFCA) approach, with a better balance between model complexity in terms of Multi-adds (14.13G) and performance. The proposed model utilizes cascading connections to enable the learning of both low and high-frequency feature maps. The proposed RFCA method consists of Multi-Feature Aggregation (MFA), Frequency Content Awareness (FCA), and Feature Consolidation (FC) block. FCA block's purpose is to boost the high-frequency contents. The extracted feature map obtained by the MFA and FCA blocks is consolidated by the Feature Consolidation (FC) Block. This approach greatly aids in learning from high-level complex features. Visual results and quantitative metrics of PSNR and SSIM exhibit the accuracy of the proposed approach on synthetic benchmark super-resolution datasets. The experimental analysis shows that the proposed approach outperforms other existing methods for SISR in terms of memory footprint and visual quality.