Compression of multilevel signals

Claim 2

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted valve and the value of the signal, a method for encoding the prediction errors, comprising the step of: (a) assigning each prediction error to one of a plurality of predetermined error ranges, each error range being a function of the sign and the number of significant digits of the prediction error value; (b) assigning a probability distribution to each such error range; and (c) encoding a prediction error using the probability of a corresponding error range to which the prediction error is assigned.

Claim 3

A method according to claim 1 including the step of encoding the absolute value of the prediction value with its assigned error range.

Claim 4

A method according to claim 1 including the steps of assinging to an error range designated as corrresponding to a specific error value, prediction errors having a corresponding number of significant digits up to a predetermined value, and assigning to a separate error range all error values having a number of significant digits greater than said predetermined value.

Claim 5

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, comprising the steps of: (a) converting each prediction error into an exponent part and a mantissa part; (b) encoding the exponent part statistically, using a subset of prediction errors calculated from others of the plurality of multilevel signals as context; and (c) combining the encoded exponent part with an encoding of the mantissa part.

Claim 6

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, comprising the steps of: (a) converting each prediction error into an exponent part and a mantissa part; (b) encoding the exponent part statistically, using a subset of prediction errors calculated from others of the plurality of multilevel signals as context; and (c) combining the encoded exponent part with an encoding of the mantissa part.

Claim 7

A method according to claim 4 wherein the step of converting the prediction error into an exponent part and a mantissa part is performed using a base of two.

Claim 8

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, comprising the steps of: (a) if the prediction error is zero, then encoding a zero; and (b) if the prediction error is not zero, then (i) encoding the sign of the prediction error; (ii) converting the magnitude of the prediction error into an exponent part and a mantissa part; (iii) encoding the exponenet part statistically, using a subset of prediction errors calculated from others of the plurality of multilevel signals as context; and (iv) combining the encoded exponent part with an encoding of the mantissa part.

Claim 9

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, comprising the steps of: (a) if the prediction error is zero, then encoding a zero; and (b) if the prediction error is not zero, then (i) encoding the sign of the prediction error; (ii) converting the magnitude of the prediction error into an exponent part and a mantissa part; (iii) encoding the exponenet part statistically, using a subset of prediction errors calculated from others of the plurality of multilevel signals as context; and (iv) combining the encoded exponent part with an encoding of the mantissa part.

Claim 10

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, an apparatus for encoding the prediction errors comprising: (a) means for converting each prediction error into an exponent part and a maintissa part; (b) means for storing a predetermined subset of said converted exponent parts and mantissa parts; (c) means for generating a context from said stored exponent parts; and (d) means for encoding the exponent part and the mantissa part, the exponent part being encoded statistically using said context from the context generator.

Claim 11

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, an apparatus for encoding the prediction errors comprising: (a) means for converting each prediction error into an exponent part and a maintissa part; (b) means for storing a predetermined subset of said converted exponent parts and mantissa parts; (c) means for generating a context from said stored exponent parts; and (d) means for encoding the exponent part and the mantissa part, the exponent part being encoded statistically using said context from the context generator.

Claim 12

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, an apparatus for encoding the prediction errors comprising: (a) means for determining whether the prediction erorr is zero; (b) means coupled to said determining means and responsive to a prediction error of zero for encoding a zero; and (c) means coupled to said determining means and responsive to a prediction error other than zero for generating; (i) an indication of the sign of the prediction error; (ii) an indication of the digit position of the most significant non-zero digit of the absolute value of the prediction error; (iii) an indication of the absolute value of the digits following said most significant non-zero digit; and (d) means responsive to said indications from said generator (c) for encoding a non-zero.

Claim 13

In the compression of multilevel signals by conditionally predicting each original signal and encoding the prediction error, said prediction error being calculated as a function of the difference between the predicted value and the actual value of the signal, a method for encoding the prediction errors, an apparatus for encoding the prediction errors comprising: (a) means for determining whether the prediction erorr is zero; (b) means coupled to said determining means and responsive to a prediction error of zero for encoding a zero; and (c) means coupled to said determining means and responsive to a prediction error other than zero for generating; (i) an indication of the sign of the prediction error; (ii) an indication of the digit position of the most significant non-zero digit of the absolute value of the prediction error; (iii) an indication of the absolute value of the digits following said most significant non-zero digit; and (d) means responsive to said indications from said generator (c) for encoding a non-zero.