You are on page 1of 3

RGB and Geobody Delineation

Fundamental to obtaining information from spectral decomposition techniques for seismic


interpretation is the ability to examine and compare the response at different frequency bands. RGB
(Red-Green-Blue) blending of 2D images has been used extensively for this purpose in other domains
and also in spectral decomposition of seismic data (Hall 2004). RGB volume visualisation and
opacity rendering of 3D seismic spectral decomposition data has not been reported previously.

RGB blending combines the information provided by 3 input images to generate a single “full colour”
output image. An RGB image can be considered to utilise a 3D colour space, C, with distance along
the axes representing the relative intensity of each of the three colours Red, Green and Blue (Figure
1). An RGB image is formed by selecting a colour based on a position within the 3D colour
space whose coordinates are defined by the values in the three input images.

Cout(x,y,z) = C(IR(x,y,z), IG(x,y,z), IB(x,y,z)) (2)

Where Cout(x,y,z) is the colour assigned to the point (x,y,z) in the output image and IR(x,y,z), IG(x,y,z)
and IB(x,y,z) are the voxel values at point (x,y,z) in the input images that are assigned to control the
Red, Green and Blue contributions respectively.

Volumetric RGB blended images provide a tool for assessing rapidly and easily the results of the
spectral decomposition (Figure 2). However, once a feature of interest has been identified using the
spectral decomposition information it is useful to be able to use that information to produce a 3D
Geobody that delineates the feature.

Figure 1 Red green blue (RGB) color space


Figure 2 Volume RGB blend showing the different seismic response (different colour) in different
fault blocks.

Loop Reconvolution
Many methods can enhance-frequency bandwidth of conventional seismic data that assist interpreters
in solving particular problems. These methods range from an additional deconvolution step through
wavelet extractions to ‘black box’ proprietary techniques. We present a very simple frequency
optimized loop reconvolution method (FOX) which can be implemented quickly and repeatedly (for
optimizing visual preferences), and appears to offer many of the same cosmetic benefits as the
commercially available techniques.

Conventionally processed poststack seismic data has a comparatively narrow frequency spectrum with
the signal utilising only a fraction of the available bandwidth up to the Nyquist. High-frequency
enhancements aim to sharpen the data, ostensibly defining structures and pinch-outs more clearly.
Irrespective of whether these techniques actually recover any missing or hidden information from the
data, they can help with the interpretation because events can appear more sharply defined and are less
swamped by the low frequency ringing that characterises conventional seismic data. Interpreters
conceptualise the dataset as a sequence of discrete spatially located geological events, whereas data
processors emphasise the spatially distributed wavefield and its associated spectral properties. We
have developed a process that makes the data appear more like a sequence of discrete horizons with a
virtually unlimited bandwidth. The method consists of 4 basic steps:
1. Oversample the input SEGY data.
2. Generate a new sparse spike reflectivity series weighted by the interpolated amplitudes at all of the
maxima and minima locations.
3. Convolve the resulting reflectivity series with a suitably high frequency wavelet. A broad
bandwidth Klauder or other wavelet with smaller sidelobes gives a good result. Visual preferences
will determine the optimum frequency band for that dataset.
4. Optional spatial filtering of the final section gives a smoother appearance yet preserves the
discontinuities that we are trying to visually enhance.
Reference
Hall. M. and Trouillot, E. (2004) Predicting stratigraphy with spectral decomposition: Canadian
Society of Exploration Geophysics, Expanded Abstracts

P. Young and A. Wild. (2005) Cosmetic Enhancement of Seismic Data by Loop Reconvolution.
Canadian Society of Exploration Geophysics National Convention p.78-90