Integral transforms, data compression, and automatic analysis of seismic sections

It is shown that seismic sections contain 10⁵ times more data than is needed to fully describe a completely interpreted section. Three ways of reducing the size of a data set are suggested: mapping into subsets of smaller dimension; normalization providing a reduction factor of four or more; and conversion into zero crossings, further reducing the data by a factor of 20. If combined, these methods can reduce the data set by a factor of 100 or more and may also improve the S/N (signal/noise) of the data and their effective dynamic range. In implementing these methods, two families of transforms that are designed to extract the information from the data (integral and morphological transforms) are introduced. Interpretation by visual inspection is more of a pattern classification rather than quantified analysis of the data. Ways to identify such patterns using clustering algorithms are suggested