Method for reducing the quantity of data in image coding

Claim 1

In a method for reducing the quantity of image data to be transmitted using image coding including the steps of subdividing each image into matrix shaped partial image regions of NXN pixels, discretely transforming the image data of each of said partial image regions from a pixel domain into a spectral domain to produce spectral coefficients, classifying each of said partial image regions, using said spectral coefficients, so that each of said partial image regions is characterized by one of predesignated ordered classes, and utilizing the class characterization for the partial image regions to determine which spectral coefficients must be transmitted; the improvement wherein said step of classifying comprises: forming absolute values from the spectral coefficients of a partial image region; comparing the absolute values of said spectral coefficients, value for value, with previously determined corresponding visual discernibility thresholds to provide a logic 0 if a spectral coefficient is less than its corresponding visual discernibility threshold and a logic 1 if a spectral coefficient is greater or equal to its corresponding visual discernibility threshold; intermediately storing the binary comparison results of said step of comparing in a first shift register as a binary bit sequence pattern of logic 0's and logic 1's representing said spectral coefficients of said a partial image region; using an operator on each binary bit of said sequence pattern stored in said first shift register to suppress singular binary bits representing said spectral coefficients and to obtain a modified binary sequence pattern of logic 0's and logic 1's representing the partial image region; storing said modified binary bit sequence pattern in a further shift register; by means of a logical term gate array connected to said further shift register, forming output signals corresponding to term variables of a predetermined set of terms from said modified binary bit sequence pattern, with said terms having been selected as minimum quantities from which all preceding of said ordered classes can be easily compiled and distinguished; sequentially generating binary signals corresponding to said ordered classes by means of a class decoder; comparing the output signals from the class decoder and from the logical term gate array with one another until a first one of the ordered classes has been found which is congruent with a term formed by the logical term gate array; and identifying said first one of the ordered classes by a class code which is forwarded in this form for further processing.