An Adjustable Gradient Filter for Volume Visualization Image Enhancement

This paper presents an improved method for computation of the gradient from volume data, which can be used to improve the rendering of small details in images of this data. A common method for estimation of the gradient vector is the center difference, which also acts as a fixed-response low pass filter, smoothing details in the image. The center difference method is extended in this paper to allow control of sensitivity to fine details in the source data. The amount of smoothing performed by the gradient computation can be tuned to filter out high frequency noise and aliased energy while retaining as much as possible of the valid frequency content which may be lost using the conventional center difference. Volume visualization techniques extract and display structures from three dimensional arrays of sample values representing physical objects. The shading of surfaces of the structures to simulate natural lighting makes the features of these structures visible in images. An important step in the image rendering process is the determination of the orientation of an extracted surface in the form of a gradient vector at each point on the surface. This gradient vector is then used as a surface normal vector for shading calculations. In addition, some volume visualization techniques also use the gradient vector for classification of structures contained in the volume.