Retrieval of text information from natural scene images and video frames is achallenging task due to its inherent problems like complex character shapes,low resolution, background noise, etc. Available OCR systems often fail toretrieve such information in scene/video frames. Keyword spotting, analternative way to retrieve information, performs efficient text searching insuch scenarios. However, current word spotting techniques in scene/video imagesare script-specific and they are mainly developed for Latin script. This paperpresents a novel word spotting framework using dynamic shape coding for textretrieval in natural scene image and video frames. The framework is designed tosearch query keyword from multiple scripts with the help of on-the-flyscript-wise keyword generation for the corresponding script. We have used atwo-stage word spotting approach using Hidden Markov Model (HMM) to detect thetranslated keyword in a given text line by identifying the script of the line.A novel unsupervised dynamic shape coding based scheme has been used to groupsimilar shape characters to avoid confusion and to improve text alignment.Next, the hypotheses locations are verified to improve retrieval performance.To evaluate the proposed system for searching keyword from natural scene imageand video frames, we have considered two popular Indic scripts such as Bangla(Bengali) and Devanagari along with English. Inspired by the zone-wiserecognition approach in Indic scripts[1], zone-wise text information has beenused to improve the traditional word spotting performance in Indic scripts. Forour experiment, a dataset consisting of images of different scenes and videoframes of English, Bangla and Devanagari scripts were considered. The resultsobtained showed the effectiveness of our proposed word spotting approach.
展开▼
