Back to results
US 11,388,416 B2Granted
Video compression using deep generative models
AssigneeQUALCOMM Incorporated
InventorsAmirhossein Habibian, Taco Sebastiaan Cohen
Life of this patent
Estimating the term…
Abstract
Certain aspects of the present disclosure are directed to methods and apparatus for compressing video content using deep generative models. One example method generally includes receiving video content for compression. The received video content is generally encoded into a latent code space through an auto-encoder, which may be implemented by a first artificial neural network. A compressed version of the encoded video content is generally generated through a trained probabilistic model, which may be implemented by a second artificial neural network, and output for transmission.
Claims30
A method for compressing video, comprising:
- receiving video content for compression;
- encoding the received video content into a latent code space through an auto-encoder implemented by a first artificial neural network to generate an encoded video content, wherein the auto-encoder was trained by:
- generating a reconstructed version of a first video content by decoding an encoded received first video content;
- and
- comparing the reconstructed version of the first video content to the received first video content;
- generating a compressed version of the encoded video content through a probabilistic model implemented by a second artificial neural network different from the first neural network, wherein the trained probabilistic model comprises an auto-regressive model of a probability distribution over four-dimensional tensors;
- and
- outputting the compressed version of the encoded video content for transmission.
5 more elements
A system for compressing video, comprising:
- at least one processor configured to:
- receive video content for compression;
- encode the received video content into a latent code space through an auto-encoder implemented by a first artificial neural network configured to execute on the at least one processor, wherein the auto-encoder was trained by:
- generating a reconstructed version of a first video content by decoding an encoded received first video content;
- and
- comparing the reconstructed version of the first video content to the received first video content;
- generate a compressed version of the encoded video content through a probabilistic model implemented by a second artificial neural network configured to execute on the at least one processor, wherein the trained probabilistic model comprises an auto-regressive model of a probability distribution over four-dimensional tensors;
- and
- output the compressed version of the encoded video content for transmission;
- and
- a memory coupled to the at least one processor.
8 more elements
A method for decompressing encoded video, comprising:
- receiving a compressed version of an encoded video content;
- decompressing the compressed version of the encoded video content based on a probabilistic model implemented by a first artificial neural network into a latent code space, wherein the probabilistic model comprises an auto-regressive model of a probability distribution over four-dimensional tensors;
- decoding the encoded video content out of the latent code space through an auto-encoder implemented by a second artificial neural network, wherein the auto-encoder was trained by:
- generating a reconstructed version of a first video content by decoding an encoded first video content;
- and
- comparing the reconstructed version of the first video content to the received first video content;
- and
- outputting the decoded video content for display.
5 more elements
A system for decompressing encoded video, comprising:
- at least one processor configured to:
- receive a compressed version of an encoded video content;
- decompress the compressed version of the encoded video content into a latent code space based on a probabilistic model implemented by a first artificial neural network configured to execute on the at least one processor, wherein the probabilistic model comprises an auto-regressive model of a probability distribution over four-dimensional tensors;
- decode the encoded video content out of the latent code space through an auto-encoder implemented by a second artificial neural network configured to execute on the at least one processor, wherein the auto-encoder was trained by
- generating a reconstructed version of a first video content by decoding an encoded first video content;
- and
- comparing the reconstructed version of the first video content to the received first video content;
- and
- output the decoded video content for display;
- and
- a memory coupled to the at least one processor.
8 more elements
Dependencies are inferred from the claim text.