Motion compensating coder employing an image coding control method

An image coder employing motion-compensation techniques for coding input signals having multiple textures, comprising:

• means for coding input signals including
• means for generating a prediction signal of the input signal,
• means, connected to said generating means, for combining the input signal and said prediction signal,
• first circuitry connected to said combining means and including means for transforming and means for quantizing said combined signals, said transforming means generating coefficients representing said combined signal, and
• second circuitry connected to said first circuitry and including means for encoding and means for buffering said transformed and quantized signal, said buffering means generating buffer occupancy information;
• and
• a control section connected to said quantizer means and said buffering means and including means for generating a quantizer step size based upon said buffer occupancy information received from said buffering means and the texture of the input signal and for applying said quantizer step size to said quantizing means wherein said control section includes
• means, responsive to said input signal, for transforming the input signal and generating coefficients representing the input signal,
• a variance processor which receives and processes said coefficients from said input signal transforming means and coefficients from said combined signal transforming means and produces a first variance for said coefficients of said input signal and a second variance for said coefficients of said combined signal, and
• a coding controller connected to said variance processor, said buffering means, and said quantizing means and which includes
• means for determining at least one bit allocation for the input signal,
• means for generating at least one quantizer step size for the input image, and
• means for adjusting said quantizer step size to control a coding rate of the input signal.

A method of coding an input signal having multiple textures and being divisible into a plurality of segments, comprising the steps of:

• generating a prediction signal of the input signal;
• combining the input signal and said prediction signal to form a combined signal;
• transforming said combined signal to produce coefficients of said combined signal;
• quantizing said combined signal;
• coding and buffering said quantized combined signal in a buffer;
• tracking buffer occupancy of said buffer;
• transforming the input signal to produce coefficients of said input signal;
• determining the texture of the segments of the input signal by processing said coefficients of said combined signal and said coefficients of said input signal and producing a first variance and a second variance, respectively for each segment;
• allocating a number of bits for coding each segment of the input image based upon said first and second variances corresponding to each segment and a number of bits pre-allocated for coding the input signal;
• assigning a quantizer step size for each segment based upon said allocated bits in said allocating step for the segment;
• recursively adjusting said quantizer step size of said assigning step based upon said tracked buffer occupancy;
• and
• controlling said quantizing step based upon said quantizer step size from said adjusting step.

A method of controlling a coding rate of a motion compensating coder, the coder being responsive to an input, having a prediction signal and a pre-allocated number of bits for coding an input signal and tracking occupancy of a buffer holding coded signals, the input signal having multiple textures and being divisible into a plurality of segments, comprising the steps of:

• transforming a combined signal of the input signal and the prediction signal to produce coefficients of said combined signal;
• transforming the input signal to produce coefficients of said input signal;
• determining the texture of the segments of the input signal by processing said coefficients of said combined signal and said coefficients of said input signal and producing a first variance and a second variance, respectively for each segment;
• allocating a number of bits for coding each segment of the input signal based upon said first and second variances corresponding to each segment and the number of bits pre-allocated for coding the input signal;
• assigning a quantizer step size for each segment based upon the allocated bits in said allocating step for the segment;
• recursively adjusting the quantizer step sizes of said assigning step based upon the buffer occupancy;
• and
• controlling the coding rate at the coder by inputting thereto the adjusted quantizer step sizes.