The top-left subsurface time-slice in Figure 5 displays a buried river channel that meanders diagonally from the South-West to the North-East corner. The time-slice possibly contains an entire channel system besides the single visible river channel. The buried river channel is embedded in gently dipping beds that generally strike South-East to North-West and are cut by the channel's sediments. The vertical cross-sections illustrate the gentle reflector dip.
The bottom row of Figure 5 shows a coherency image by CTC The CTC coherency time-slice clearly delineates the edge of the river. In addition to the major river channel, the CTC coherency time-slice displays additional discontinuities which are not obvious in the original image. These events could be spurious events from other time-slices or could indicate an underlying channel system that is not visible in the original time-slice. Overall, the image has an almost binary character in which locations are flagged as being part of an edge or of an homogeneous reflector volume.
The Lomoplan outputs at the bottom of Figure 6 successfully reject the image's smooth reflectors and delineate the major channel. Unlike the CTC coherency time-slice, the Lomoplan image does not indicate the channel's edge points uniformly. The edges are highlighted by an elusive light-shadow effect that corresponds to the filter's prediction directions. Some isolated strong values in the original data result in some strong-amplitude, high-frequency residuals. Unfortunately, the application of a spatial low-pass filter to the Lomoplan output failed to suppress these events without deteriorating the image of the channel significantly.
Lomoplan does not show the complex channel system that the coherency data shows. This might demonstrate Lomoplan's superior resolution or its inferior sensitivity depending on whether one believes in the existence of the channel system or not.
To highlight the differences between the Lomoplan images and the CTC coherency image, we computed the amplitude spectrum of the various time-slices. Figure 7 displays the original time-slice images in the top row and the corresponding amplitude spectra in the bottom row. Each spectrum is normalized to a total energy of 1 and all spectra are clipped with the same value.
The original time-slice image has a low frequency power spectrum (see Figure 7). The Lomoplan East-West prediction error filter removes the predictable East-West energy and causes a vertical low energy band in the spectral plot (see bottom center plot of Figure 7). The splitting and merging of the original image to process individual subcubes is not noticeable in the power spectrum. Finally, the power spectrum of the CTC coherency time-slice displays two distinct components: it contains a low frequency component which is similar to the amplitude spectrum of the original subsurface image. Additionally, the spectrum contains a modulated, cross-shaped component along the zero frequency axes. The low frequency component surprised us, since we expected the coherency cube to whiten the spectrum.