A study on the restoration of mural painting images in Fujian palace temples based on improved recurrent generative adversarial networks

Purpose A sample block-based improved algorithm for mural image restoration is proposed for the restoration of precious Fujian palace and temple mural art. Methodologies The paper adopts image restoration algorithm with recurrent generative adversarial network and multiscale fusion coordinated attention mechanism,firstly,add the same mapping loss to the recurrent consistency loss,then improve the coordinated attention mechanism and propose the multiscale fusion coordinated attention mechanism,and finally,introduce the multiscale fusion coordinated attention mechanism in the generator to improve the coordinatedness of the generated image. Event Compared with CycleGAN traditional algorithm,the image restoration algorithm based on CycleGAN improved recurrent generative adversarial network and multi-scale fusion coordinated attention mechanism has excellent effect on restoring the mural painting images of Fujian Palace Temple. Conclude Image restoration algorithms with recurrent generative adversarial networks and multi-scale fusion of coordinated attention mechanisms can provide good restoration of faded Quanzhou Shenggonggong Palace mural images without relying on expert guidance.