Ancient Book Seals Segmentation Based on Automatic Region Growing

Seal is one of the components of ancient Chinese books, paintings or documents. Sealing is usually the last step before the final completion of works such as calligraphy or painting. The seal is of great significance in the study of ancient books, it indicates the copyright, ownership and other information. In this paper, due to the color characteristics of the seal, an improved region growing algorithm and contour detection method have been proposed that used for the seal segmentation. Firstly, the image that includes seals has been converted to Lab color space and obtain specific pixels as the initial seeds for region growing method. After that, the contour detection operation has been performed on the result image of the region growing algorithm to get the candidate position of the seals. Finally, the division of the seal extracted has been realized. The results of experiments showed that the segmentation accuracy of up to 98%.