Restoration of oracle bone inscriptions using a fast residual shrinkage denoising network with fractal gradient

To address the low-texture and strong-edge characteristics of oracle bone inscription images, this research proposes a restoration method based on a fast residual shrinkage denoising network integrated with adaptive box-counting fractal gradient features. The adaptive differential box counting fractal gradient, when employed as a loss function, can be dynamically adjusted according to the local characteristics of the image, thereby enabling more accurate gradient estimation and enhanced representation of strong edge features. By integrating sampling pairs and soft thresholding mechanisms into the residual denoising network, the model achieves improved denoising performance while preserving edge details, thus accomplishing image restoration tasks more effectively. Compared with traditional approaches, the method achieves superior performance in both subjective visual quality and objective evaluation metrics, substantially improving edge clarity in oracle bone inscriptions.