03-05-2011, 09:34 AM
[attachment=13270]
Abstract
In this paper, a multi-ROIs medical imagecompression algorithm with edge feature preserving isproposed. By taking advantage of characteristics ofmedical image and human eye, the algorithm extractsimage edge information by using canny operator first,and then divides image into regions-of-interest (ROIs)and non-ROIs. To ROIs, lossless compressionalgorithm JPEG2000 is used. To non-ROIs, highcompression ratio algorithm SPIHT is used. Theexperimental results show that our algorithm hasfollowing advantages: (1) The algorithm has highcompression ratio as well as keeping the quality ofROIs; (2) The algorithm can deal with edge Gibbseffect and improve the quality of image reconstruction;(3) The algorithm is adaptive and practical, which canbe used for remote medical image compression,storage and transfer.
Key Words: Medical image compression; SPIHT;JPEG2000; multi-ROIs; Edge Feature Preserving
1. Introduction
Different kinds of medical images have differentcharacters. But all of them have a common feature, i.e.useful information is usually gathered and occupied asmall area in the image. This feature can be utilized tocompress an image. To useful areas (which is called asregions-of-interest, i.e. ROI), compress lowcompression ratio or even without compression to keepthe quality of image in this area. To other areas,compress with high compression ratio to make highdensity compression. The compressed image is keptwith useful information as well as small size.In traditional medical image compression algorithm,an image is processed by wavelet transform first, andthen transform coefficients are selected for differentareas to keep lossless compression in the area. Thismethod is suggested by JPEG2000 standard. Butusually after the transform, the edge of imagereconstruction is obscure if the bit ratio is low, whichis called as Gibbs effect [1]. To deal with this problem,an algorithm which combines edge detection and ROIcoding is proposed. In this algorithm, the edge ofimage is detected by using Canny edge detector [2]first. Then ROIs are selected. Different compressionmethods are used to ROIs and non-ROIs. The result bitstreams are integrated to obtain the compressed image.The experimental results show that the proposedalgorithm not only keeps a high compression ratio butalso gets a good image reconstruction.
2. Original SPIHT algorithm
The wavelet-based image encoding can improve thecompression rate and the visual quality considerably,and many researchers propose different methods forencoding the wavelet-based images. The SPIHT [3]algorithm is an efficient method for lossy and losslesscoding of natural images. The SPIHT algorithm adoptsa hierarchical quad-tree [4, 5] data structure onwavelet-transformed image. The energy of a wavelettransformedimage is concentrated on the lowfrequency coefficients. A tree structure, called asspatial orientation tree (SOT), naturally defines thespatial relationship of the hierarchical pyramid. Fig. 1shows how a spatial orientation tree is defined in apyramid constructed with recursive four sub-bandssplitting. The coefficients are ordered in hierarchies.According to this relationship, the SPIHT algorithmsaves many bits that specify insignificant coefficients.
Abstract
In this paper, a multi-ROIs medical imagecompression algorithm with edge feature preserving isproposed. By taking advantage of characteristics ofmedical image and human eye, the algorithm extractsimage edge information by using canny operator first,and then divides image into regions-of-interest (ROIs)and non-ROIs. To ROIs, lossless compressionalgorithm JPEG2000 is used. To non-ROIs, highcompression ratio algorithm SPIHT is used. Theexperimental results show that our algorithm hasfollowing advantages: (1) The algorithm has highcompression ratio as well as keeping the quality ofROIs; (2) The algorithm can deal with edge Gibbseffect and improve the quality of image reconstruction;(3) The algorithm is adaptive and practical, which canbe used for remote medical image compression,storage and transfer.
Key Words: Medical image compression; SPIHT;JPEG2000; multi-ROIs; Edge Feature Preserving
1. Introduction
Different kinds of medical images have differentcharacters. But all of them have a common feature, i.e.useful information is usually gathered and occupied asmall area in the image. This feature can be utilized tocompress an image. To useful areas (which is called asregions-of-interest, i.e. ROI), compress lowcompression ratio or even without compression to keepthe quality of image in this area. To other areas,compress with high compression ratio to make highdensity compression. The compressed image is keptwith useful information as well as small size.In traditional medical image compression algorithm,an image is processed by wavelet transform first, andthen transform coefficients are selected for differentareas to keep lossless compression in the area. Thismethod is suggested by JPEG2000 standard. Butusually after the transform, the edge of imagereconstruction is obscure if the bit ratio is low, whichis called as Gibbs effect [1]. To deal with this problem,an algorithm which combines edge detection and ROIcoding is proposed. In this algorithm, the edge ofimage is detected by using Canny edge detector [2]first. Then ROIs are selected. Different compressionmethods are used to ROIs and non-ROIs. The result bitstreams are integrated to obtain the compressed image.The experimental results show that the proposedalgorithm not only keeps a high compression ratio butalso gets a good image reconstruction.
2. Original SPIHT algorithm
The wavelet-based image encoding can improve thecompression rate and the visual quality considerably,and many researchers propose different methods forencoding the wavelet-based images. The SPIHT [3]algorithm is an efficient method for lossy and losslesscoding of natural images. The SPIHT algorithm adoptsa hierarchical quad-tree [4, 5] data structure onwavelet-transformed image. The energy of a wavelettransformedimage is concentrated on the lowfrequency coefficients. A tree structure, called asspatial orientation tree (SOT), naturally defines thespatial relationship of the hierarchical pyramid. Fig. 1shows how a spatial orientation tree is defined in apyramid constructed with recursive four sub-bandssplitting. The coefficients are ordered in hierarchies.According to this relationship, the SPIHT algorithmsaves many bits that specify insignificant coefficients.