PT Unknown
AU Partha Pratim Roy
TI Multi-Oriented and Multi-Scaled Text Character Analysis and Recognition in Graphical Documents and their  Applications to Document Image Retrieval
PY 2010
AB With the advent research of Document Image Analysis and Recognition (DIAR), animportant line of research is explored on indexing and retrieval of graphics rich documents. It aims at finding relevant documents relying on segmentation and recognitionof text and graphics components underlying in non-standard layout where commercialOCRs can not be applied due to complexity. This thesis is focused towards text information extraction approaches in graphical documents and retrieval of such documentsusing text information.Automatic text recognition in graphical documents (map, engineering drawing,etc.) involves many challenges because text characters are usually printed in multioriented and multi-scale way along with different graphical objects. Text charactersare used to annotate the graphical curve lines and hence, many times they followcurvi-linear paths too. For OCR of such documents, individual text lines and theircorresponding words/characters need to be extracted.For recognition of multi-font, multi-scale and multi-oriented characters, we haveproposed a feature descriptor for character shape using angular information from contour pixels to take care of the invariance nature. To improve the efficiency of OCR, anapproach towards the segmentation of multi-oriented touching strings into individualcharacters is also discussed. Convex hull based background information is used tosegment a touching string into possible primitive segments and later these primitivesegments are merged to get optimum segmentation using dynamic programming. Toovercome the touching/overlapping problem of text with graphical lines, a characterspotting approach using SIFT and skeleton information is included. Afterwards, wepropose a novel method to extract individual curvi-linear text lines using the foreground and background information of the characters of the text and a water reservoirconcept is used to utilize the background information.We have also formulated the methodologies for graphical document retrieval applications using query words and seals. The retrieval approaches are performed usingrecognition results of individual components in the document. Given a query text,the system extracts positional knowledge from the query word and uses the same togenerate hypothetical locations in the document. Indexing of documents is also performed based on automatic detection of seals from documents containing clutteredbackground. A seal is characterized by scale and rotation invariant spatial featuredescriptors computed from labelled text characters and a concept based on the Generalized Hough Transform is used to locate the seal in documents.
ER