Adaptive video coding method

Claim 1

A method of adaptively encoding a sequence of frames of image information, wherein at least some of said frames contain a plurality of video objects, for providing a compressed video signal to a transmission channel by means of a buffer having a variable input bit rate and a substantially constant output bit rate comprising the steps of: encoding each of said video objects in each of a set of frames using coding means including a processor for performing discrete cosine transform to produce transform coefficients and a quantizer for quantizing the transform coefficients to generate image-representative code bits at a variable rate, said encoding step producing texture, motion and shape information for each said video object; storing said image representative code bits in said buffer; restricting the contents of said buffer with respect to a predetermined limit value by adjusting quantization parameters utilized by said quantizer with respect to a reference value according to a quadratic rate distortion model to increase or decrease the amount of code bits generated by said coding means for said video objects in successive ones of said frames; estimating a target number of bits for encoding each video object in each successive frame in a sequence occurring over a predetermined time interval following the first frame by distributing a target number of bits for all objects in each video object plane among said objects in accordance with a function of relative motion, size and variance parameters associated with corresponding objects in the corresponding object plane; and setting said variable rate for encoding at one of at least a higher rate and a lower rate to avoid overflow of said buffer while preserving image quality.

Claim 2

A method of adaptively encoding a sequence of frames of image information, wherein at least some of said frames contain a plurality of video objects, for providing a compressed video signal to a transmission channel by means of a buffer having a variable input bit rate and a substantially constant output bit rate comprising the steps of: encoding each of said video objects in each of a set of frames using coding means including a processor for performing discrete cosine transform to produce transform coefficients and a quantizer for quantizing the transform coefficients to generate image-representative code bits at a variable rate, said encoding step producing texture, motion and shape information for each said video object; storing said image representative code bits in said buffer; restricting the contents of said buffer with respect to a predetermined limit value by adjusting quantization parameters utilized by said quantizer with respect to a reference value according to a quadratic rate distortion model to increase or decrease the amount of code bits generated by said coding means for said video objects in successive ones of said frames; estimating a target number of bits for encoding each video object in each successive frame in a sequence occurring over a predetermined time interval following the first frame by distributing a target number of bits for all objects in each video object plane among said objects in accordance with a function of relative motion, size and variance parameters associated with corresponding objects in the corresponding object plane; and setting said variable rate for encoding at one of at least a higher rate and a lower rate to avoid overflow of said buffer while preserving image quality.

Claim 15

A method of adaptively encoding a sequence of frames of image information, wherein at least some of said frames contain a plurality of video objects, for providing a compressed video signal to a transmission channel by means of a buffer having a variable input bit rate and a substantially constant output bit rate comprising the steps of: encoding each of said video objects in each of a set of frames using coding means including a processor for performing discrete cosine transform to produce transform coefficients and a quantizer for quantizing the transform coefficients to generate image-representative code bits at a variable rate, said encoding step producing texture, motion and shape information for each said video object; storing said image representative code bits in said buffer; restricting the contents of said buffer with respect to a predetermined limit value by adjusting quantization parameters utilized by said quantizer with respect to a reference value according to a quadratic rate distortion model to increase or decrease the amount of code bits generated by said coding means for said video objects in successive ones of said frames; estimating a target number of bits for encoding each video object in each successive frame in a sequence occurring over a predetermined time interval following the first frame by distributing a target number of bits for all objects in each video object plane among said objects in accordance with a function of relative motion, size and variance parameters associated with corresponding objects in the corresponding object plane; and coding said shape information for each object according to a mask; size converting each macroblock of each said object for encoding according to a predetermined conversion ratio; reconstructing the original size of each said macroblock; determining a conversion error for each pixel block within said macro block; comparing said conversion errors to a predetermined threshold to identify error pixel blocks; and increasing said conversion ratio and redetermining conversion errors and comparison thereof to said threshold until said threshold is not exceeded or until a maximum conversion ratio is reached.

Claim 16

A method of adaptively encoding a sequence of frames of image information, wherein at least some of said frames contain a plurality of video objects, for providing a compressed video signal to a transmission channel by means of a buffer having a variable input bit rate and a substantially constant output bit rate comprising the steps of: encoding each of said video objects in each of a set of frames using coding means including a processor for performing discrete cosine transform to produce transform coefficients and a quantizer for quantizing the transform coefficients to generate image-representative code bits at a variable rate, said encoding step producing texture, motion and shape information for each said video object; storing said image representative code bits in said buffer; restricting the contents of said buffer with respect to a predetermined limit value by adjusting quantization parameters utilized by said quantizer with respect to a reference value according to a quadratic rate distortion model to increase or decrease the amount of code bits generated by said coding means for said video objects in successive ones of said frames; estimating a target number of bits for encoding each video object in each successive frame in a sequence occurring over a predetermined time interval following the first frame by distributing a target number of bits for all objects in each video object plane among said objects in accordance with a function of relative motion, size and variance parameters associated with corresponding objects in the corresponding object plane; and coding said shape information for each object according to a mask; size converting each macroblock of each said object for encoding according to a predetermined conversion ratio; reconstructing the original size of each said macroblock; determining a conversion error for each pixel block within said macro block; comparing said conversion errors to a predetermined threshold to identify error pixel blocks; and increasing said conversion ratio and redetermining conversion errors and comparison thereof to said threshold until said threshold is not exceeded or until a maximum conversion ratio is reached.

Claim 22

A method of adaptively encoding a sequence of frames of image information, wherein at least some of said frames contain a plurality of video objects, for providing a compressed video signal to a transmission channel by means of a buffer having a variable input bit rate and a substantially constant output bit rate comprising the steps of: encoding each of said video objects in each of a set of frames using coding means including a processor for performing discrete cosine transform to produce transform coefficients and a quantizer for quantizing the transform coefficients to generate image-representative code bits at a variable rate, said encoding step producing texture, motion and shape information for each said video object; storing said image representative code bits in said buffer; restricting the contents of said buffer with respect to a predetermined limit value by adjusting quantization parameters utilized by said quantizer with respect to a reference value according to a quadratic rate distortion model to increase or decrease the amount of code bits generated by said coding means for said video objects in successive ones of said frames; setting said variable rate for encoding at one of at least a higher rate and a lower rate to avoid overflow of said buffer while preserving image quality; size converting said shape information for each macroblock of each said object according to a predetermined conversion ratio; determining a conversion error for each pixel block within each said macro block; comparing said conversion errors to a predetermined Alpha threshold to identify error pixel blocks; increasing said conversion ratio and redetermining conversion errors and comparison thereof to said Alpha threshold until said Alpha threshold is not exceeded or until a maximum conversion ratio is reached; skipping the coding of a frame for a frame period whenever the difference between buffer bit capacity and current buffer level is less than a predetermined margin at the end of the encoding of all video objects in a frame; said setting of said variable rate being determined by counting a number of consecutive skipped frames in a time interval immediately preceding said setting step, said variable rate being set at said higher rate when said number of skipped frames is less than a predetermined number and being set at said lower rate when said number of skipped frames is equal to or greater than said predetermined number; after encoding, determining whether said higher rate or said lower rate is operative; and increasing said Alpha threshold if said lower rate is operative and decreasing said Alpha threshold if said higher rate is operative for a succeeding coding interval.

Claim 23

A method of adaptively encoding a sequence of frames of image information, wherein at least some of said frames contain a plurality of video objects, for providing a compressed video signal to a transmission channel by means of a buffer having a variable input bit rate and a substantially constant output bit rate comprising the steps of: encoding each of said video objects in each of a set of frames using coding means including a processor for performing discrete cosine transform to produce transform coefficients and a quantizer for quantizing the transform coefficients to generate image-representative code bits at a variable rate, said encoding step producing texture, motion and shape information for each said video object; storing said image representative code bits in said buffer; restricting the contents of said buffer with respect to a predetermined limit value by adjusting quantization parameters utilized by said quantizer with respect to a reference value according to a quadratic rate distortion model to increase or decrease the amount of code bits generated by said coding means for said video objects in successive ones of said frames; setting said variable rate for encoding at one of at least a higher rate and a lower rate to avoid overflow of said buffer while preserving image quality; size converting said shape information for each macroblock of each said object according to a predetermined conversion ratio; determining a conversion error for each pixel block within each said macro block; comparing said conversion errors to a predetermined Alpha threshold to identify error pixel blocks; increasing said conversion ratio and redetermining conversion errors and comparison thereof to said Alpha threshold until said Alpha threshold is not exceeded or until a maximum conversion ratio is reached; skipping the coding of a frame for a frame period whenever the difference between buffer bit capacity and current buffer level is less than a predetermined margin at the end of the encoding of all video objects in a frame; said setting of said variable rate being determined by counting a number of consecutive skipped frames in a time interval immediately preceding said setting step, said variable rate being set at said higher rate when said number of skipped frames is less than a predetermined number and being set at said lower rate when said number of skipped frames is equal to or greater than said predetermined number; after encoding, determining whether said higher rate or said lower rate is operative; and increasing said Alpha threshold if said lower rate is operative and decreasing said Alpha threshold if said higher rate is operative for a succeeding coding interval.