Abstract: This study explores the use of Scale Invariant Feature Transform (SIFT), Modified-Scale Invariant Feature Transform (M-SIFT) for feature extraction of text, typically the punjabi text, however, this area is mature enough for languages like English, Chinese and Arabic. This study presents an ongoing effort to build fusion based text extraction approach using SIFT also variations in SIFT are experimented and it is denoted as M-SIFT. Fused image is obtained using principal component analysis based fusion method. From within the fused image, region of interest is perceived to find the part of frame which contains text in it. Region, thus, supposed is cropped for feature extraction. Feature extraction is carried out by using SIFT and M-SIFT based algorithm separately. This study addresses the problems in the various stages of the development of a recognition system for Punjabi text in news ticker present in Punjabi news videos.
Pooja and Renu Dhir, 2017. News Ticker Recognition: Performance Analysis SIFT and M-SIFT. Journal of Engineering and Applied Sciences, 12: 9483-9489.