High-frequency art image resolution reconstruction based on HFRA-SRGAN algorithm: A study on image quality metrics and super-resolution GANs

This study focuses on the problem of insufficient clarity and pixels in current art images, and constructs an art image resolution reconstruction model based on high-frequency reconstruction and amplification with super-resolution generative adversarial networks (HFRA-SRGAN). The new model enhances the network model's ability to process image data by adding attention networks and residual blocks, enhancing the algorithm's ability to analyze art images and the clarity of images. The research results show that the use of the new algorithm can improve the peak signal-to-noise ratio and algorithm accuracy of art images. The performance of the new algorithm and the initial algorithm has better image parsing ability compared to current advanced algorithms. At 4x magnification, the peak signal-to-noise ratio of the HFRA-SRGAN model is 24.31 dB and the structural similarity is 0.632, which are 3.07 dB and 0.090 higher than those of the SRGAN model and 1.78 dB and 0.035 higher than those of the Super-Resolution Convolutional Neural Network (SRCNN) model, respectively. Among them, the new algorithm's reconstruction recognition accuracy is about 2% higher than the SRGAN model, and is about 1% higher than the image super-resolution reconstruction algorithm model. This study demonstrates the potential enhancement of parsing and reconstruction capabilities in high-frequency art images, thereby providing a valuable reference point for future research in this domain.