GB1488912A - Method and apparatus for encoding an information signal and compressing the transmission bandwidth - Google Patents

Abstract

1488912 Data transmission; facsimile systems RCA CORPORATION 11 Dec 1974 [26 Dec 1973] 53542/74 Heading H4P A method of coding a two-dimensional image comprises generating a first matrix of n Î n signals representing an information signal and from this generating a second matrix derived from the first comprising 2n x 2n signals having even symmetry with respect to X-Y axes, i.e the first matrix forming one of four quadrants each of which is a mirror image with the adjacent quadrant in each direction, and two dimensional transform coding the second matrix to provide a plurality of signals corresponding to the transform coefficients which represent the information signal. A Fourier transform may be used equivalent to a series expansion so that each sub picture is represented as though it was a period of a two dimensional periodic function. An n x n (i.e. 4 Î 4) matrix produces 16 complex coefficients which require 16 numbers to specify, also there may be discontinuities in period functions. Reforming a 4 x 4 matrix into an 8 x 8 matrix, each 4 Î 4 edge section being a mirror image of the adjacent 4 Î 4 along each quadrant defining axis, avoids this as the 8 Î 8 matrix has even symmetry, requires only 16 numbers to specify the sub picture of 64 Fourier coefficients, also avoids discontinuities (see Figs. 5 and 7, not shown). The apparatus comprises scanner 20 of any type, e.g. vidicon tube, laser, facsimile scanner or scanned matrix array, discharging digital signals representing 4 rows alternately into buffers 22, 24 read out also alternately into data expander 26 which converts 4 Î 4 arrays into 8 Î 8 even symmetric matrices. In FFT network 28, an appropriate two dimensional Fourier transform is computed and passed to block 30 which thins and quantizes input signals before transmission by any line or radio medium. At the receiver (Fig. 9, not shown) incoming signals are decoded to reconstruct the Fourier transforms which are then passed to FFT network (44) performing an inverse transform to provide an 8 x 8 matrix to 46 where it is reformed into a 4 x 4 matrix for display. The arrangement provides steadily rising and falling representation of the information which avoids discontinuities hence reducing dissipation of high-frequency energy. Haelamand or Karhunen-LoÚve transforms may be alternatively used.