Optimization algorithm for 3D image visual communication based on digital image reconstruction

Existing three-dimensional (3D) reconstruction methods have deficiencies in image depth estimation accuracy, texture mapping continuity, and illumination consistency, which lead to obstacles such as geometric distortion and texture fracture in the visual communication of 3D images, affecting the efficiency of information acquisition and immersive experience. To solve these problems, this paper proposes a 3D image visual communication optimization algorithm that integrates neural implicit modeling and a multi-scale visual perception mechanism. By jointly encoding the image depth map and Red-Green-Blue (RGB) map into the Neural Radiance Fields(NeRF) voxel hashing network, the continuity of spatial structure expression and the integrity of texture restoration are improved. Structural similarity constraints and perceptual consistency loss functions are introduced to enhance visual stability and subjective quality under different viewing angles, and context completion and detail enhancement of edge texture missing areas are achieved through graph neural networks. User evaluation results show that this method can shorten the average recognition time by up to 33.1% in all target recognition tasks, improve the average subjective immersion score by up to 58.2%, and reduce the Root Mean Square Error (RMSE) of depth reconstruction in occluded areas to 0.164 meters. The Structural Similarity Index Measure(SSIM) of high-frequency texture areas reached 0.871, and the Learned Perceptual Image Patch Similarity(LPIPS) was stable at 0.162 under ±45° viewing angle offset, which effectively improved the image's structural restoration quality and functional visual communication performance, and provided algorithm support for high-precision virtual expression in multiple scenarios.