This study aims to enhance the overall balance among image detail restoration, structure preservation, and model adaptability in image restoration tasks. The study proposes an optimized Pulse Coupled Neural Network (PCNN) image restoration model integrated with artificial intelligence (AI), leveraging the bio-inspired advantages of PCNN in image processing. In terms of model construction, a PCNN-AI restoration system with end-to-end training capability is designed. The PCNN backbone network extracts temporal pulse features of images. The AI module adaptively optimizes key parameters such as coupling coefficients and firing thresholds. The experimental design uses the DIV2K dataset to construct diverse image degradation samples. Comparative results show that the optimized model has significant advantages in image quality: For urban architectural images, it achieves a Peak Signal-to-Noise Ratio (PSNR) of 34.07, Structural Similarity Index (SSIM) of 0.957, and Learned Perceptual Image Patch Similarity (LPIPS) of 0.036. The optimized model outperforms models like Swin Transformer for Image Restoration (SwinIR), Nonlinear Activation Free Network (NAFNet) and Restoration Transformer (Restormer). In simulation experiments, the model demonstrates strong robustness and interpretability, with a structural salience matching degree of 0.94 and semantic consistency projection rate of 0.97, which is significantly superiors to comparative model. Thus, this study provides a fusion path for image restoration and understanding that balances structural modeling, biological mechanisms, and AI optimization. It offers theoretical value and engineering application significance for algorithm design and system deployment in image restoration tasks.
Zhang et al. (2026) studied this question.