Encoding device and method for utility-driven video compression

1 . An encoding device for utility-driven video compression, the encoding device comprising circuitry configured to:

accept an input video having a first data volume comprising at least one frame, each frame comprising a plurality of blocks;

identify at least one region of interest in a first spatial region of the first data volume, the first spatial region comprising a first plurality of included blocks;

identify at least a feature reflecting a structural or content attribute of the at least one region of interest;

identify a region of exclusion identified as a region containing blocks associated with at least one object detected in the input video, the at least one object including at least one of a human face, a person, a vehicle, or a license plate, the at least one object being a feature to be selectively excluded from the bitstream;

generate an output video, wherein:

the output video contains a second data volume that is less than the first data volume; and

the output video preserves the at least a feature of the region of interest; and

generate an encoded bitstream including the output video and signaling information including a header indicating a block-level association of blocks to regions in the second data volume.

2 . The encoding device of claim 1 , wherein the encoding device is further configured to accept the input video by:

receiving an input bitstream; and

decoding the input video from the input bitstream.

3 . The encoder of claim 1 , further configured to identify the at least a feature of interest by:

receiving at least a supervised annotation indicating the at least a feature of interest; and

identifying the at least a feature of interest using the at least a supervised annotation.

4 . The encoder of claim 1 , further configured to identify the at least a feature of interest using a neural network.

5 . The encoder of claim 4 , further configured to:

receive an output bitstream recipient characteristic; and

select the neural network from a plurality of neural networks as a function of the output bitstream recipient characteristic.

6 . The encoding device of claim 1 , wherein the at least a feature of interest includes at least one of an audio feature, a visual feature, and an element of metadata.

7 . The encoding device of claim 1 , wherein the header includes a privacy flag for each block and the flag for the blocks of a region of exclusion are set to indicate exclusion from the encoded bitstream.

8 . The encoding device of claim 1 , wherein the-block-level association includes a spatial region identifier.

9 . The encoding device of claim 1 , wherein encoding the bitstream further comprises compressing the output video.

10 . The encoding device of claim 1 , wherein the header is a slice header.

11 . The encoding device of claim 1 , wherein the signaling information is provided in a sequence parameter set (SPS) in a slice header of the bitstream.

12 . The encoding device of claim 1 , wherein the feature describes at least one of spatial and temporal characteristics of a frame or group of frames in the output video.

13 . The encoding device of claim 1 , wherein signaling information is provided as supplemental information in the bitstream.

14 . The encoding device of claim 13 , wherein the second feature is motion information of blocks sufficient to perform motion analysis.

15 . A method for utility-driven video compression, the method comprising:

accepting, by an encoding device, an input video having a first data volume comprising at least one frame, each frame comprising a plurality of blocks;

identify, by the encoding device, at least one region of interest in a first spatial region of the first data volume, the first spatial region comprising a first plurality of included blocks;

identifying, by the encoding device, at least a feature reflecting a structural or content attribute of the at least one region of interest;

identifying, by the encoding device, a region of exclusion identified as a region containing blocks associated with at least one object detected in the input video, the at least one object including at least one of a human face, a person, a vehicle, or a license plate, the at least one object being a feature to be selectively excluded from the bitstream;

generating, by the encoding device, an output video, wherein:

the output video contains a second data volume that is less than the first data volume; and

the output video preserves the at least a feature of the region of interest; and

generating, by the encoding device, an encoded bitstream including the output video and signaling information including a header indicating a block-level association of blocks to regions in the second data volume.

16 . The method of claim 15 , wherein accepting the input video further comprises:

receiving an input bitstream; and

decoding the input video from the input bitstream.

17 . The encoder of claim 15 , wherein identifying the at least a feature of interest further comprises:

receiving at least a supervised annotation indicating the at least a feature of interest; and

identifying the at least a feature of interest using the at least a supervised annotation.

18 . The encoder of claim 15 , wherein identifying the at least a feature of interest further comprises identifying the at least a feature of interest using a neural network.

19 . The encoder of claim 18 further comprising:

receiving an output bitstream recipient characteristic; and

selecting the neural network from a plurality of neural networks as a function of the output bitstream recipient characteristic.

20 . The method of claim 15 , wherein the at least a feature of interest includes at least one of an audio feature, a visual feature, and an element of metadata.

21 . The method of claim 15 , wherein the header includes a privacy flag for each block and the flag for the blocks of a region of exclusion are set to indicate exclusion from the encoded bitstream.

22 . The method of claim 15 , wherein the block-level association includes a spatial region identifier.

23 . The method of claim 15 , wherein encoding the bitstream further comprises compressing the output video.

24 . An encoding device for utility-driven video compression, the encoding device comprising circuitry configured to:

accept an input video comprising a plurality of frames and having a first data volume;

identify in the input video at least a first feature relevant to a first application and a second feature relevant to a second application, wherein the first and second feature each concerning a specific structural or content attribute of the input video;

generate a first encoded application specific sub-bitstream for human viewing consisting essentially of a subset of frames including the first feature and having a volume less than the first data volume; and

generate a second encoded application specific sub-bitstream for a machine-based application consisting essentially of a subset of frames including the second feature and having a volume less than the first data volume, the second encoded application specific sub-bitstream comprising motion information of blocks in the input video and excluding texture data and color data of the blocks, the second encoded application specific sub-bitstream being incapable of reconstructing pixels of the input video necessary for human viewing.