Video Text Extraction Algorithm Based On Harris And Color Features

In face of plentiful video, people urgently need a kind of tool that can availably inspect and obtain video contents. The video text includes abundant high-level semantic information. It is important to exact texts from the video image and identify them for video browsing and searching.This paper adopts an improved video text detection method to process text detection stage and gets the location of text region. Text image is obtained by text extraction based on text color, which is according to cluster color of corner neighboring pixels. Firstly, it needs to fulfill the Harris corners detection in the original region, and judge whether the video frame is text frame or not according to the user-defined multi-character threshold value. After detecting the former three text frames, it needs to ensure the candidate text region to complete text location. Then, the obtained corners are calculated by use of clustering the color of corners and neighboring pixels in the candidate text region. It utilizes the color class of former three text frames to get the final text color. Finally, It needs to obtain the pixel by use of the text color. This algorithm processes text-effect enhancement is essential to decrease the phenomenon of text rupture after completing text extraction. The experiment indicates that candidate text region can lessen the range of text examination and the algorithm is better for text detection under the complicated back ground. Moreover, it can relatively ensure the text by use of the text color information, which is obtained by corners and neighboring pixels.