Existing lossless coding models yield only up to 3:1 compression. However, a much higher lossless compression can be achieved for certain medical images when the images are segmented prior to applying integer to integer wavelet transform and lossless coding. The methodology used in this research work is to apply a contour detection scheme to segment the image first. The segmented image is then wavelet transformed with integer to integer mapping to obtain a lower weighted entropy than the original. An adaptive arithmetic model is then applied to code the transformed image losslessly. For the male visible human color image set, the overall average lossless compression using the above scheme is around 10:1 whereas the compression ratio of an individual slice can be as high as 16:1. The achievable compression ratio depends on the actual bit rate of the segmented images attained by lossless coding as well as the compression obtainable from segmentation alone. The computational time required by the entire process is fast enough for application on large medical images.
|Proceedings of SPIE - The International Society for Optical Engineering
|Published - 2000
|Medical Imaging 2000: Image Processing - San Diego, CA, USA
Duration: Feb 14 2000 → Feb 17 2000