Method for fully adaptive calibration of a prediction error coder

A method for adaptive calibration of a prediction error coder on a computing processor, the method comprising:

• a first step of initialization, comprising:
• a first initialization of a histogram of values for counting occurrences of individual low values and of groups of large values, allocating storage or memory resources for the first initialization, wherein if the storage or memory resources for the first initialization are not already available, setting all histogram bins to zero occurrences and mapping each bin of the histogram to a given value for low values, or range of values for high values, to be compressed;
• a second initialization of a variable block size to a given value;
• and
• an allocation of computing and memory resources for using the prediction error coder;
• a second step of reception and accumulation of block-size values, wherein for each received value, one more occurrence is summed to the histogram bin associated to that value, storing the received value in a buffer and checking whether the number of accumulated values has reached the block size value;
• a third step of analysis of the histogram and determination of the coding option comprising:
• checking if the histogram accumulates a sufficient number of occurrences in the first bins corresponding to the lowest values, if so, a low entropy coding option is selected for the prediction error coder;
• if the low entropy conditions are not fulfilled, new thresholds of accumulated number of occurrences are checked for the first, wherein if said new low entropy conditions are fulfilled, a double-smoothed coding option is selected for the prediction error coder;
• and
• finally, if the double-smoothed entropy conditions are not fulfilled, the large coding option is selected for the prediction error coder;
• a fourth step of analysis of the histogram and determination of a coding table, selecting the number of bits to be used for each of the prediction error coder coding segments that fulfill a given condition depending on the selected coding option of the third step, for the relation between the accumulated occurrences of each range of values and the required bits to code the range of values;
• a fifth step of output to a header with the prediction error coder coding option determined in the third step and the coding table determined in the fourth step, wherein, for each value stored in the buffer, said value stored in the buffer is coded with the prediction error coder allocated in the first step using the determined coding option and the determined coding table, outputting the resulting code;
• and
• wherein previous steps are repeated if more values need to be compressed.