Document Image Retrieval from Incomplete Queries Using Texture Features

Abstract

Document image retrieval (DIR) is an important part of many documentimage processing systems such as paperless office systems, digitallibraries and so on. It helps the users to find out the most similar documentimages from a document image database. Most of the researches havebeen carried out with complete queries which were present in thedatabase, but in many cases distorted or incomplete images can beencountered. This distortion or incompetence is due to some missinginformation, some undesirable objects, blurring, noise due to documentprinting, scanning etc. This paper describes an approach for retrieval ofincomplete and distorted document images based on visual features usingtexture information for retrieval from large document image database. AGray Level Co-occurrence Matrix (GLCM) features for texture analysiswere proposed and provide a comprehensive experimental evaluation.