Video OCR: indexing digital news libraries by recognition of superimposed captions

Abstract

The automatic extraction and recognition of news captions and annotations can be of great help locating topics of interest in digital news video libraries. To achieve this goal, we present a technique, called Video OCR (Optical Character Reader), which detects, extracts, and reads text areas in digital video data. In this paper, we address problems, describe the method by which Video OCR operates, and suggest applications for its use in digital news archives. To solve two problems of character recognition for videos, low-resolution characters and extremely complex backgrounds, we apply an interpolation filter, multiframe integration and character extraction filters. Character segmentation is performed by a recognition-based segmentation method, and intermediate character recognition results are used to improve the segmentation. We also include a method for locating text areas using text-like properties and the use of a language-based postprocessing technique to increase word recognition rates. The overall recognition results are satisfactory for use in news indexing. Performing Video OCR on news video and combining its results with other video understanding techniques will improve the overall understanding of the news video content.

Semantic Based Sport Video Browsing

“…The text detection methods based on this characteristic assume that text regions have uniform colors and satisfy certain constraints on size, shape, and spatial layout. The second is the texture alike characteristic of the text regions [65,33,66,67]. The text detection methods based on texture information usually assume that the text regions have special texture patterns.…”

Annotating News Video with Locations

“…Additional locations are mapped from "organization" terms/phrases with self-contained locations, such as "Capitol Hill", using a manually created mapping list. Note that location terms are sometimes superimposed on the video frames, which can be recognized by video optical character recognition (VOCR) techniques [7]. However, the VOCR output tends to be errorful on low-resolution news video, and they offer few distinct locations since most of them overlap with those from transcript.…”

“…While we choose not to rely on the errorful locations recognized by VOCR [7] (Section 2), they are nevertheless useful due to their similarity to the true location terms. In Fig.4, for example, Iraq is recognized as Lraq, differing by only one character.…”

A localization/verification scheme for finding text in images and video frames based on contrast independent features and machine learning methods

“…(18). This verification scheme removed 7255 regions from the 7537 false alarm regions while only reject 23 true text lines, which gives a 99.76% RRR and a 97% RPR as listed in Table 4.…”

“…Temporal information is usually helpful for detecting captions since they are mostly stationary but not very helpful for the scene text, which may have various motion and transforms. Sato [18] and Lienhart [12] computed the maximum or minimum value at each pixel position over frames. The values of the background pixels that are assumed to have more variance through video sequence will be pushed to black or white while the values of the text pixels are kept.…”