Damaged Area Detection And Segmentation For Thangka Image Combined With Domain Knowledge

At present, digital image restoration has become a research topic in the field of digital image processing. There are a wide range of applications in protection of cultural relics, old movies and photo fix and so on. Because of time, climate and keeping improper, Thangka sustained different degrees of damage. Digital image restoration provides a strong protection for Thangka art heritage. But in most of the repair work the damaged regions need to be marked by the professional personnel, or indiscriminately do the same with all pixels to do the same. It is unfavorable to do large quantities of image processing. So before the repair, the damaged area must be separated.There is a challenging research focus in the fields of computer and pattern recognition. It is image segmentation. Its goal is to find area in an image which is coherent in semantic. In this topic, the regions which have semantic consistency are the damaged regions in Thangka image.Because the types of damaged regions on Thangka image varied, according to the characteristics of the different types of damage area, the types are divided in to two using SVM classifier. They are the spots and the non-spots. The non-spots includes the simple block, the threshold change obviously and the complicated. For a damaged Thangka Image to be segmented, firstly determine it belongs to which damage type and the according this to choice segmentation method. For the simple block, under the RGB color space using region growing method to segment, growth rule is based on color similarity coefficient. For the regions whose threshold change obviously, use iterative OTSU segmentation method. For the spots, combine the maximum entropy and morphology to segment them. For the complicated, considering Thangka image color features and texture features, combine with color features and texture features, fuzzy rate and FCM to segment the damaged area.Finally use the segmentation method studied in this paper to do some experiments. And we made comparative analysis on the experimental results from the subjective and objective aspects. The experimental results show that the different segmentation method according to different types of damaged area, has obtained the good experimental effect.