D CT and MR Volume Data Sets

The performance of a filter set also depends on image characteristics. The smoothness, sharpness, and noise of an image affect the compression performance of a filter set [10]. For a three-dimensional image set, the image characteristics may differ within 2D slices and in the interslice direction. A uniform 3D wavelet may not provide the best performance because the correlation of pixels within a slice is different from that between slices. Typically, the distance of adjacent pixels within a slice varies from 0.3 to 1.2 mm, whereas the distance between slices can be much higher. Each slice has a certain thickness and two adjacent slices are separated by a gap. We define the slice distance to be the sum of the slice gap and slice thickness. Typically, the slice distance varies from 1 to 10 mm for CT images and from 1.5 to 8 mm for MR images. Hence, the optimal wavelet filter set within slices (in the x and y directions) may not be the same as the optimal wavelet filter set in the slice (z) direction.

Image characteristics such as resolution, noise, and sharpness are determined by anatomy and image modality. Tables 1

TABLE 1 Selected CT image sets for evaluating wavelet filters


Pixel size (mm)

Pixel size (mm)


0 0

Post a comment