Efficient Chain-Code Encoding for Segmentation-Based Image Compression

Abstract

This paper pmsmn.s a new and efticient method of encoding oniform image mgions and lines. Regions and lines are obtained as the result of image segmentation. split and merge image compression, or as the output of lii and polygon drawing algorithms. Lines and contours of uniform regions. am enccded using chain-code. The chaincode is obtained in a way that is &icient with respect to bit-rate and produces lossless contour and line encoding. A lossy method for contour encoding is also presented. A set of experiments to compare the performamx of traditional chain-cadc contour enwding with the improved contour encoding is presented. The results show a reduction of about 50% in the bit-rate with no reconst~ction error. Efficient lossless chain-encoding of region boundaries Simple modifications to the enwding can lead to a signiticant saving in bit rate. PavMis* proposes a relative chaincode with variable code per direction. The code used in this paper is a fixed length relative code with priority where a lower number denotes higher priority. This coding methcd is designed to imxase. code redundancy. A cut of almost 50% in the chain-code can be a&ieved by modifying the traditional approach of contour truing into a region boundary tracing method. The problem with cantour following is that pixels on region contours are also on the boundaries between regioos. Hence, the cootour following algtithm enccdes each boundary twice. The change in approach entails that chains does not describe closed objects. Themfom. a few adjustment to the enccdiug scheme am requhd. Fit. one can not use the condition that the ccurdiuates of the initial pixel and the coordinates of theterminatingpivella~chain~ssame,asthemeanstoidentifytheterminatiooofachainRather.achainis ended with a relative. prioritized code denoting going back to the pm&us chain element. Second. this approach has au additional overhead due to the mquiremat to include the coordinates of the initial pixel. Now. the same region may be represented by mom dun one chain and hence mom than a initial pixel. Another modification is that instead of attxhing the information about a region to it chain-xde, the information about all the regions in the image proceeds or succeeds the chain-codes of the regions. The coding scheme has a built-in LZW compressia~ In a fcrk point, this algorithm checks the compression table. If possible it chooses a route that is an eney in the table in favor of a route that requims updating the table. A lossy option that produce smooth edges is also available Experiments and results The different improvement3 described in this paper has been applied to four images. One of these images ha-s been obtained from line and polygon drawing. Other images are “‘natural” images that have been passed through segmentation. Our tidings are: 1) Unix compression of the regular chain-cede achieves between 1.5 times compression to 10 times compression. 2) Boundary Tracing with lmsless compmssion is better than contour tracing with lossless compression. It’s effect is two times better than Unix wmpressicm. 3) Implementing a built-in LZW has a slight improvement in the compression. 4) The. lossy compression increases the bit-rate without noticeable distortion. Conclusion The results of these experiments show au improvement of about 48% (average over the four images) in the size of the cede with no additional distortion. Compressim with distmtion can braease the compression while adding objective and subjective distortion. We have found that lossy compressico of the code at 46% (average over the four images) compression has negligible negative effect on the. subjective image quality. 1. Computer Science Program, Florida Institute of Techaokgy, Melbomne, FL 32901; tami@cs.fit.edu 2. S L. Horowitz and T. Pavlidis, “Pictme segmentation by a tree traversal algorithm”. Joournnl of the association for Computer Machinery, vol. 23. no. 2. April 1976: 368-388, 1068.0314/96$5.00

DOI: 10.1109/DCC.1996.488387

Cite this paper

@inproceedings{Tamir1996EfficientCE, title={Efficient Chain-Code Encoding for Segmentation-Based Image Compression}, author={Dan E. Tamir and Kim Phillips and Abdul-razzak Abdul-karim}, booktitle={Data Compression Conference}, year={1996} }