Research On The Priority Extraction Of Salient Regions With Different Cultural Elements In The Dunhuang Grotto Murals

Dunhuang murals have occurred discoloration problem, thus digitizationprotection of them is imminent. The study is based on this background, we extract asignificant area of Dunhuang murals to filter out information that we are notinterested in, which can largely reduce computational complexity. Currently imagesaliency detection method is mostly based on feature comparison of pixels, instead ofregions, there are also not priority extraction of salient regions that we can notmeasure how important the region is. In this thesis, we study to extract significantregions of Dunhuang murals and prioritize them, it can extract most important regionsof the images to some extent, and classify in a significant degree between theseregions.This thesis focuses on research of the priority extraction of salient regions withdifferent cultural elements in the Dunhuang Murals. Its main contents are as follows:(1) Compared with other segment method, we choose Mean Shift algorithm foimage segmentation of Dunhuang Murals. First of all,we compared the thresholdsegmentation method, based on segmentation and edge-based segmentation methodson the advantages and disadvantages of image segmentation, then we introduce andchoose the image segmentation method based on Mean Shift algorithm.Finally, usingthis method for image processing of Dunhuang murals.(2) Designed and proposed a new calculation method of significant region value.We analyzed some features which can present whether a region is significant or not,the features include color features, shape features, location features. Then we choosesome representative features in them, to get a formula which used to calculatesignificant value of regions.(3) Designed and proposed a transfer method of significant region in theDunhuang Murals.After calculating the significant value of salient regions, we selectthe region with the largest significant value as beginning of the transfer process, thenwe use the transfer function to identify significant degree lower priority than a largerone, until reaching the end of the condition, we will find all significant regions of a picture, and mark their priority. The main role of the transfer function is to identifythe priority of the division image significantly, to simulate the human visual transfermechanism to some extent.(4) Realized significant prioritization extraction of Dunhuang murals. Afterimage segmentation of Dunhuang murals, significant value was calculated for eachdivided region, according transfer method to get the priority extraction of salientregions with different cultural elements in Dunhuang murals.In this thesis, we have researched on the priority extraction of salient regions withdifferent cultural elements in the Dunhuang Grotto Murals. Under the circumstancesof the complexity of huge data of Dunhuang murals, we can analyze these structuralvariation patterns over time better, and provide a scientific basis for Dunhuangfrescoes multivariate data retrieval.