A self-alignment node matching method for ceramic artifact texture misalignment combining feature weighted filtering and random sampling matching

Addressing the issues of fragmented distribution and misaligned texture features in ceramic artifacts. We propose an image stitching method that integrates feature-weighted filtering with random sampling matching to achieve precise filtering and stitching of ceramic artifact images. The global pixel grayscale distribution of the image is statistically analyzed. Adaptive weights are assigned to each histogram dimension based on the grayscale distribution characteristics of ceramic textures. A weighted feature vector space is constructed. Weighted Euclidean distance is used to measure the similarity between feature vectors and quantify the matching degree. Develop a grayscale distribution difference index table to enable target image retrieval. Define descriptor regions to construct high-dimensional feature descriptors. Iteratively solve for the optimal homography matrix using random sample consistency, achieving precise and seamless stitching of ceramic artifacts. The assembled ceramic artifact images achieved a peak signal-to-noise ratio of 65.41 and a mean squared error of 0.0187. This approach effectively resolves issues of fragmented distribution and misaligned texture features in ceramic artifacts, significantly enhancing the efficiency and accuracy of ceramic artifact image assembly.