« Back

Video-Search Technique Relies on Text

Searching video efficiently and effectively has long been a research challenge. Some techniques use text found in video—such as street signs or building names to find a specific frame or section of a video. However, recognizing text can be difficult because it could be occluded or could vary in size or orientation. A research team from the National University of Singapore and the A*STAR Institute for Infocomm Research developed a multistep automatic text-recognition process. The software uses image masks, which are used to isolate a portion of an image, to enhance the contrast between text and background, then employs an algorithm to find the text. The software then uses a technique called boundary growing to grow a text box that surrounds the line of text in the video to verify that the content of the box is indeed text. The researchers say they are trying to improve their software’s accuracy and ability to work with different types of text. They published their findings in IEEE Transactions on Circuits and Systems for Video Technology. (PhysOrg)(IEEE Transactions on Circuits and Systems for Video Technology)

Trackback URL: