Scanned Compound Document Image Compression using Optimization Techniques

In Existing systems, compound document compression follow a top down approach that first segments the original image into three distinct layers ,the background layer containing smooth objects, the foreground layer containing information regarding text and graphics colors and mask layer by using the MRC model. These distinct layers are then compressed individually by various encoders like JPEG2000, H.264/AVC Intra, and Digipaper, MMP algorithm to compress the document. In this paper, the existing scheme uses a block based classification that separates pictorial macro-blocks in to text and graphics ones and the macro-blocks are signalled using a binary mask. Each of these two types of macro-blocks are then encoded using a version of the Multidimensional Multiscale Parser (MMP) algorithm specifically designed to encode it. The non-smooth blocks (text/graphics) are compressed using MMP-text algorithm and the smooth blocks are compressed by means of MMP-FP (flexible Partition (FP) algorithm. In proposed scheme, Macro blocks are segmented with the use of morphological filters and its separated blocks are using R-D optimization techniques. The Modified MMP algorithms are implemented to efficiently encode the text blocks and the image blocks. For encoding text blocks Modified MMP-Text algorithm is used, where the binary tree based segmentation process is implemented. For encoding the image blocks, Modified MMP-FP algorithm is implemented, where Quad-tree based segmentation process is carried out. Modified MMP has an advantage that it is not sensitive to pixel classification errors. Also it increases the compression ratio, improves the compression effectiveness, improves the image quality, reduces the search space between the pixels and also minimizes the computational complexity.