Tamil text detection in videos using Gradient Vector Flow and Fuzzy C-Means

Abstract

In videos, detecting text with multifaceted scenarios is perplexing. Texts in those videos have content full data facts that will be applied for various applications. Here, a system is proposed to enrich the text detection process from video. Here, a new method is implemented that detects Tamil text based on Gradient Vector Flow (GVF) and fuzzy c-means. First the video is split into number of frames. To circumvent temporal redundancy in each frame, a key frame is chosen and the frame where the text be located is identified to be the key frame. The dominant edge pixel is identified in that frame by the sobel edge map. Edge components are detected conforming towards the dominant pixel in sobel detector for constructing Text Candidates (TC).Clustering of a pixel is performed to detect text by using fuzzy c means clustering algorithm. Finally text is detected.