期刊
IET IMAGE PROCESSING
卷 7, 期 7, 页码 686-693出版社
INST ENGINEERING TECHNOLOGY-IET
DOI: 10.1049/iet-ipr.2012.0041
关键词
biomedical MRI; data compression; fractals; image coding; medical image processing; compression ratio; quasi-lossless fractal coding algorithms; quasi-lossless fractal compression method; quasi-lossless fractal coding scheme; machine learning-based model; encoding computation time; fractal transformations; domain blocks; encoding time; peak signal-to-noise ratio; compression ratio; MRI; compress magnetic resonance images; quasi-lossless fractal coding; standard fractal coding; fractal-based coding algorithms; medical image compression; fractal coding methods
In this study, the performance of fractal-based coding algorithms such as standard fractal coding, quasi-lossless fractal coding and improved quasi-lossless fractal coding are evaluated by investigating their ability to compress magnetic resonance images (MRIs) based on compression ratio, peak signal-to-noise ratio and encoding time. For this purpose, MRI head scan test sets of 512 x 512 pixels have been used. A novel quasi-lossless fractal coding scheme, which preserves important feature-rich portions of the medical image, such as domain blocks and generates the remaining part of the image from it, has been proposed using fractal transformations. One of the biggest tasks in fractal image compression is reduction of encoding computation time. A machine learning-based model is used for reducing the encoding time and also for improving the performance of the quasi-lossless fractal coding scheme. The results show a better performance of improved quasi-lossless fractal compression method. The quasi-lossless and improved quasi-lossless fractal coding algorithms are found to outperform standard fractal coding thereby proving the possibility of using fractal-based image compression algorithms for medical image compression. The proposed algorithm allows significant reduction of encoding time and also improvement in the compression ratio.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据