US 12,087,024 B2
Image compression using normalizing flows
Abdelaziz Djelouah, Zürich (CH); Leonhard Markus Helminger, Zürich (CH); Scott Labrozzi, Cary, NC (US); Yuanyi Xue, Kensington, CA (US); Erika Varis Doggett, Los Angeles, CA (US); Jared McPhillen, Glendale, CA (US); and Christopher Richard Schroers, Zürich (CH)
Assigned to Disney Enterprises, Inc., Burbank, CA (US); and ETH Zurich, Zurich (CH)
Filed by Disney Enterprises, Inc., Burbank, CA (US); and ETH ZÜRICH, Zürich (CH)
Filed on Mar. 6, 2020, as Appl. No. 16/811,219.
Claims priority of provisional application 62/935,018, filed on Nov. 13, 2019.
Prior Publication US 2021/0142524 A1, May 13, 2021
Int. Cl. H04N 19/60 (2014.01); G06T 9/00 (2006.01); H04N 19/126 (2014.01); H04N 19/91 (2014.01)
CPC G06T 9/002 (2013.01) [H04N 19/126 (2014.11); H04N 19/60 (2014.11); H04N 19/91 (2014.11)] 16 Claims
OG exemplary drawing
 
1. An image compression system comprising:
a computing platform including a hardware processor and a system memory;
a software code including a normalizing flow stored in the system memory;
the hardware processor configured to execute the software code to:
receive an input image by the normalizing flow;
transform, by the normalizing flow serving as an alternative to an encoder, the input image to a latent space representation of the input image;
quantize the latent space representation of the input image to produce a plurality of quantized latents;
obtain a probability density function of the latent space representation of the input image based on a normalizing flow mapping of the latent space representation of the input image by the normalizing flow;
generate a bitstream using the plurality of quantized latents and the probability density function of the latent space representation of the input image; and
convert the bitstream into an output image corresponding to the input image.