Research on Image Quality Evaluation Method for Collaborative Computing of Large-Scale Image 3D Reconstruction

This paper proposes a new algorithm to effectively detect distortions and evaluate the perceptual quality of 3D reconstructed images, with a primary application context in detecting blurring, deformation, and other issues in the reconstruction of operation and maintenance images of power transmission and distribution systems. Combining the Gabor transform and convolutional neural network, distortion classification is applied to large-scale 3D reconstructed images to optimize the reconstruction process, improve efficiency, and improve classification accuracy and model robustness by handling class imbalance and distortion degree for reclassification. By optimizing the collaborative computation of large-scale 3D reconstructed images through distortion classification and correlation analysis, and utilizing rotation invariant feature descriptors and multi-resolution matching strategies, the accuracy and efficiency of image matching have been improved. A quality grading evaluation standard was established based on the descriptor association threshold to objectively quantify the quality of reconstructed images. The results indicate that the proposed method outperforms traditional methods in both accuracy and computational speed, validating the effectiveness of the method.