Decorrelating Medical Image Sets with Lifting: A New Approach.
Rahman Tashakkori, John M. Tyler, Oleg S. Pianykh, Xiaojun Qi · 2004
Abstract- In recent years, wavelet lifting is widely used in digital image processing. Different lifting schemes decorrelate digital medical image sets differently. The difference is due to the properties of the lifting schemes and the way the image data is processed by these schemes. Each medical image has a foreground (the medical data) surrounded by background. The background is not used by the radiologist for diagnosis and hence can be modified without affecting the radiology. Several new scanning techniques are used in this paper to isolate the background and to produce a one-dimensional array that contains all the original image data. This one-dimensional image array is subsequently used to decorrelate the images with wavelet lifting. The scanning technique combined with lifting provides an excellent wavelet based compression scheme. An additional compression was obtained when the image background was removed. All the images used in this paper were grayscale. Keywords: Lifting-based compression, image scanning, and image boundary treatment In the past two decades, wavelet analysis has been widely used in image processing [1][2][3]. Some wavelets decorrelate a similar set of medical images better than others do [4]. Wavelets obtained with a lifting scheme have the following four attractive properties: 1) Do not require the mechanism of Fourier transform, 2) Lead to algorithms that can be generalized to complex geometric structures, 3) Can map integers to integers producing lossless coding