Indian Journal of Science and Technology

Abstract

Storage of scanned compound documents is a challenging issue which needs an efficient method to compress the scanned document images. Many techniques are available for compressing the scanned compound images. Most of them have limitation in terms of PSNR value and compression ratio. Hence, a new compression method based on the newly announced coding paradigm called Fuzzy cluster based compression is proposed in this paper that gives high coding efficiency for an extensive variety of image types, by effectively adjusting the input image characteristics. The proposed system utilizes Fuzzy c-mean clustering, Optimization techniques and also a new Tree based compression approach called Enhanced Multidimensional MultiScale Parser (EMMP) which is used to compress the scanned compound document image effectively. Fuzzy clustering results contain a combination of accurate text/graphics and image portions. These results are given into proposed EMMP. In this work, the proposed scheme contains the Binary tree based Compression is used for non-smooth (Text/graphics) image and a quad tree based compression is used for smooth image which is greatly reducing the complexity of segmentation. Finally, pattern matching procedure has been made to match the encoded blocks based on different scale dimensions. The proposed technique provides the best performance in terms of very good PSNR value and high compression ratio. 
Keywords: Dictionary Creation, Fuzzy C-means Clustering, Pattern Matching, Scanned Compound Document, R-D Optimization