Optimization Effect of Random Sample Consistency Algorithm in Immersive 3D Animation Splicing

In order to provide viewers with a richer and more realistic visual experience and achieve the secondary utilization of 3D animation resources, an immersive 3D animation production image stitching model is designed based on Accelerated Kernelized-Adaptive-Zone-Extraction and Random Sample Consensus. The results showed that the image registration algorithm achieved the highest recall rate under changes in brightness, rotation angle, and flip angle, with values of 0.942, 0.971, and 0.831, respectively. The highest matching accuracy reached 94.17%. Meanwhile, the algorithm had the shortest execution time, with computation time for different datasets all within 400ms. The matching accuracy performed the best. The image quality processed by the designed image stitching algorithm was relatively high. The Brenner gradient function, Tenengrad gradient function, energy gradient function, and sum of modified differences 2 index simultaneously verified the clarity and detail quality of the image. Meanwhile, the peak signal-to-noise ratio exceeded the other three models, and the feature, multi-scale structure, and information content weighted similarity were all higher than 0.9. The mutual information, normalized mutual information, and normalized cross-correlation of spliced images performed the best. This study introduces new optimization strategies for the field of animation stitching, providing new ideas and methods for theoretical research in the immersive 3D animation field. This research has important practical significance for promoting the widespread application of immersive animation in film, television, gaming, and other fields.