Automatic Extraction Of Hand Writing Research On Murals Image

With the impact of digital wave in the world, some people have introduced digital images to repair technology into the mural restoration work, and have reached certain achievements. But this application to the fresco technique still did not jump out from the limitation of using artificial mask, and people still need to "tell" the computer which area should be restored. This method of manually adding mask reduces the accuracy and efficiency of the mural restoration, and need a new implementation of automatic identification of the damaged part to complete the repair work.Because of the tremendous differences among the different images, this paper firstly describes some methods which can automatically detect and identify the interesting regions of classic images. These methods used for different types of images, can obtain different extractive results.Secondly, the present study analyzes the character of different mural images, and establishes a vector based on GLCM features and locates the standard deviation value of the images. According to the distance among the vectors, we choose K average classification to classify the vectors. We have analyzed the difference between these two kinds of damaged mural images, and select the appropriate algorithm.Based on the sorted results, use image histogram-based impact detection method to detect damaged areas for one kind of the damaged murals. Then we analyze the possibility of errors happened and how to correct the error. In addition, according to the selection of the sliding window size and frame size of the writing, estimate the threshold value of impact. Moreover, we complete the repair work based on the detection.Lastly, this thesis uses classified algorithm based on connected components to extract human handwriting area for other kind of the damage murals. The approach consists of seven cascaded classifiers, each classifier focuses on one feature between hand writing connected components and the contents of connected components of murals. In order to improve the efficiency of the algorithm, we proposed a value to describe the classified efficiency, and on which to set the order of the classifiers. Thus, it finishes the repairing work based on the results of automatic extraction.