The initial results of hyperbola whitening are inconclusive, but there are some encouraging results. The spectrum of a purely kinematic hyperbola is whitened considerably after the 2-D prediction-error filter, but noise has been generated in the evanescent zone. When a hyperbola with amplitude and half-derivative filter corrections is put into the prediction-error filter, it appears that the D.C. component of the spectrum is enhanced. The noise in the evanescent zone and the high D.C. component of the spectrum may be attributed to the prediction-error filter's action on the evanescent zone. As noted above, the prediction-error filter might be modified to avoid noise in this zone.
After prediction-error filtering, each hyperbola is followed by a series of ripples. Ripples may be seen in the whitened hyperbola and in the spectrum of the whitened hyperbola. This effect may be caused by the number of filter coefficients in time being too small. Unfortunately, increasing the size of the filter may increase the computation and application costs of the process.
The noise being generated in the evanescent zone may be eliminated by constraining some elements of the prediction-error filter to be zero. Accurate control over the filter constraints may be a problem, since we are using only three coefficients in the trace direction. Controlling the evanescent-zone noise buildup is the next step in developing this process.