A new Histogram Oriented Moments descriptor for multi-oriented moving text detection in video. Issue 21 (30th November 2015)

• Record Type: Journal Article
• Authors: Khare, Vijeta; Shivakumara, Palaiahnakote; Raveendran, Paramesran
• Is Part Of: Expert systems with applications. Volume 42:Issue 21(2015)
• Page Start: 7627
• Page End: 7640
• Publication Date: 2015-11-30
• Subjects: Central moments -- Histogram Oriented Moments -- Histogram Oriented Gradients -- Video text detection -- Optical flow -- Moving caption text detection; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2015.06.002
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Histogram Oriented Moments (HOM) is proposed for text detection in videos. The HOM finds orientations using second order geometrical moments. A new hypothesis based on the dominant orientations of connected component proposed. To detect the moving text, we explore optical flow properties. Moving caption is selected based on constant velocity and uniform direction. Abstract: Developing an expert text detection system for video indexing and retrieving is a challenging task due to low resolution, complex background, non-illumination and movement of text present in a video. Besides, text detection is vital for several real time applications, such as license plate recognition, assisting a blind person and other surveillance applications. In this paper, we introduce a new descriptor called Histogram Oriented Moments (HOM) for text detection in video, which is invariant to rotation, scaling, font, and font size variations. The HOM finds orientations with the second order geometrical moments for each sliding window (overlapped block) of the input frame. The proposed method performs histogram operations on the orientations of each window to identify the dominant orientation (as a representative). Then, a new hypothesis is defined based on the dominant orientations of a connected component as the numbers of orientations, which point towards centroid of the connected components are larger than the number of dominant orientations which point away from the centroid of the connected components. The components that satisfy the above hypothesis are considered as text candidates, or else as non-text candidates. Further, to detect a moving text- we explore optical flow properties, such as velocity of text candidates to estimate the motions between temporal frames. The components which move with constant velocity and uniform direction are considered as text candidates otherwise non-text candidates. We demonstrate the proposed method's dominance over state of the art methods by testing on benchmark database, namely ICDAR 2013 and our own video datasets in terms of recall, precision and F-measure.