TEXTUAL CONTENT REMOVAL SCHEME OVER VIDEO STREAM

Abstract

This paper introduces a novel framework for Text detection and Removal. We present a two stage framework for automatic video text detection and  make the frames and remove embedded video text. Then fill-in their remaining regions by  Similar   Pixel or regions. Text locations in each frame are found via an unsupervised clustering Algorithm performed on the connected components produced by the stroke width transform. Many different types of image in-painting algorithms already in the literature. Image in-painting is filling in damaged or destroyed regions in an image with similar pixels or regions. Explore the use of the proposed approach to image text detection and perform in-painting and show promising results. Another contribution of this paper is used SWT. SWT needs an accurate edge map. The algorithms are analyzed in both theoretical ways, which have made the suitability of these image in-painting algorithms over different kinds of applications in image processing areas. We use most popular canny edge detector which advantage   from the geometric features revealed by the bandlet transform. The detected video text regions are removed, and then the video is restored by an in-painting algorithm scheme. The proposed video in-painting approach uses spatio-temporal geometric flows extracted by bandlets to recover the missing data.