A Joint Approach of Harris Corners Detection and Baseline Searching for Localization of Uyghur Text Lines in Image Sequences

This paper proposed an approach for detection of Uyghur text lines in video frames according to the writing styles and characteristics of Uyghur characters. Firstly, according to the comparative values of color sub-components between the text lines located area and background, gets the corner maps of the corresponding grayscale images by using the algorithm of Harris corner detection. Then the relatively isolated corners are removed in the corner maps by filtering, and then the corner images are fused into text regions. And finally, the edges of the grayscale images are detected by using Sobel operators, and the detected text regions are checked whether those are the true Uyghur text lines located areas or not according to the heuristic features of a true target area and the baseline feature of Uyghur text lines. The experimental results show that the proposed method achieves good effects and robustness in localization of Uyghur text lines in images sequences.