A Technique For Uyghur Text Localization In Video Images Based On Corner Intensity

Uighur as one of the four languages of the ethnic minorities in China,main use of the largest administrative region Xinjiang Uygur Autonomous Region,It is of great significance to explore and inherit the traditional national culture.However,the text is the most intuitive expression of a national culture,traditional customs,to carry forward the good spirit of the main information transfer tools.With the continuous improvement of multimedia technology in recent years,appear the images,videos and other mass media.And the more important content of information carrying the text embedded in videos or images,but also a lot of bad information through the Internet spread,so text detection and recognition of Uyghur videos/images has a great contribution to network security;the correct detection and localization of the text area are the primary condition of the text extraction and recognition system,and to locate large quantities of text areas of the video sequences requires a lot of manpower,energy and time.Therefore,the design and production of a highly efficient Uighur text area localization system has a very important research value.To solve the problem of locating the Uyghur text in video sequences,this paper proposed a new algorithm based on combining Harris corners and corners distance classification.Firstly,extract the edge of the input grayscale image,detect the Harris corners and filtered out the non text corners.Then taking horizontal projection get the segmented corners region,in this process,the position of the gap between the corners line is determined as the segmented points used stratifying corners region.Secondly,the vertical projection method is applied to each corners area,the same method is used to segment segmented corners region and determine the baseline of every stratified region,and carry on morphological expanding operations to the corners within the baseline zone get the candidate text area.And merging the connected components to determine the text box position of the candidate text area.Finally,used heuristic rules get the final positioning of the text regions.Experimental results show that the proposed method has good effect both in comparison with other similar algorithms,or in video sequences that contain a single text area or a plurality of parallel text regions,with real-time,robustness and anti-interference.