Research on Planar Mosaic Method for Colored Textures of Regularly Symmetric Jar-type Cultural Relics

Abstract. The use of stitching methods to obtain side view unfolded images of jar-shaped cultural relics with regular symmetrical shapes is of great significance for the digital presentation of cultural heritage. However, traditional image stitching methods face issues such as inaccurate feature point matching, visible seams, and color inconsistencies when processing jar-shaped cultural relics with curved surfaces from different perspectives. To address these issues, this paper proposes an improved SURF algorithm and a multi-band gradient fusion framework. The method dynamically adjusts the extraction threshold of the traditional SURF algorithm based on the image's contrast and edge density, and then extracts feature points. The RANSAC algorithm is used to filter matching points and eliminate mismatches. Subsequently, the images are fused using the multi-band gradient fusion framework, which separates low-frequency illumination information from high-frequency texture information. The low-frequency part ensures illumination consistency, while the high-frequency part repairs seam issues in the stitched areas. Experimental results show that the improved algorithm significantly increases the number of correctly matched points when processing texture images, and effectively solves the seam and color inconsistency problems in stitched images. The fused images exhibit visual effects that are significantly superior to those of traditional stitching methods. This method effectively overcomes the limitations of traditional methods in practical applications and provides strong technical support for the digital presentation of cultural relics.