GB1377594A - Method and apparatus for compressing data - Google Patents

Abstract

1377594 Compressing data SEISMOGRAPH SERVICE CORP 25 Jan 1973 [16 Feb 1972] 3888/72 Heading G4A Analogue data is converted to digital form and transformed from the time domain to the frequency domain, data values corresponding to frequencies outside a desired range being removed. As described for compressing seismic data from geophones which, after conversion to digital form, is fed either direct to a digital processor 22 (Fig. 1) or temporarily stored on magnetic tape 21 before being fed to the processor, the data is transformed either using hardware 23, or by appropriately programming the processor 22, using the Cooley-Tuckey fast Fourier transform algorithm to derive data values having real and imaginary components each of 16 bits (an appropriate number of zero data values is added to make the number of samples up to a power of two). The complete transform of the seismic data extends from the zero frequency to twice the Nyquist frequency but since the frequency domain plot is symmetrical only values up to the Nyquist frequency are retained. To prevent arithmetic overflow from occurring the input data values may be shifted, the amount of shift being stored to indicate the scaling factor between the time domain and frequency domain versions of the data. The frequency domain data is then limited to the range of interest. If for example the source used to derive the data generates a select frequency seismic waveform between 10 cps and 40 cps only data values whose frequencies lie within this range are retained. The reduced data is then cross-correlated with a frequency domain representation of the source signal by multiplying the amplitudes of corresponding frequency components and subtracting their phases from one another. This reduces the number of data values to be retained by one half. De-convolution is effected by dividing the data by a filtered version of itself to eliminate gross irregularities in the frequency response curve (in the preferred embodiment only the amplitude and not the phase of the data being modified). The 16 bit real and imaginary components of the compressed data values are then trunca ted to for example four bits since the original seismic waveform may be quite accurately reconstructed using these truncated values which are transmitted to core storage in a digital processor 27. Zero data values may be stored to replace the missing data components and a mirror image of the frequency domain data values may also be stored to extend the range back to twice the Nyquist frequency. By performing an inverse Fourier transform on the stored data the original seismic waveform may be obtained, the stored scaling factor being used to rescale the data either before or after transformation. The apparatus may also be used for compressing television or telegraphic picture scans.