US 12,439,046 B2
Variable-bit-rate image compression method and system, apparatus, terminal, and storage medium
Wenrui Dai, Shanghai (CN); Shaohui Li, Shanghai (CN); Chenglin Li, Shanghai (CN); Junni Zou, Shanghai (CN); and Hongkai Xiong, Shanghai (CN)
Assigned to SHANGHAI JIAO TONG UNIVERSITY, Shanghai (CN)
Filed by SHANGHAI JIAO TONG UNIVERSITY, Shanghai (CN)
Filed on May 23, 2025, as Appl. No. 19/218,104.
Application 19/218,104 is a continuation of application No. PCT/CN2023/126934, filed on Oct. 26, 2023.
Claims priority of application No. 202211605816.8 (CN), filed on Dec. 12, 2022.
Prior Publication US 2025/0287004 A1, Sep. 11, 2025
Int. Cl. H04N 19/124 (2014.01); H04N 19/13 (2014.01); H04N 19/147 (2014.01); H04N 19/184 (2014.01); H04N 19/30 (2014.01); H04N 19/42 (2014.01)
CPC H04N 19/124 (2014.11) [H04N 19/13 (2014.11); H04N 19/147 (2014.11); H04N 19/184 (2014.11); H04N 19/30 (2014.11); H04N 19/42 (2014.11)] 15 Claims
OG exemplary drawing
 
1. A variable-bit-rate image compression method, comprising:
forward mapping: performing forward mapping on a to-be-encoded image through a first decomposition transform neural network to obtain an initial feature map of the to-be-encoded image;
quantization: quantizing the initial feature map by a dead-zone quantizer to obtain a quantized feature map of the image;
entropy encoding: performing entropy encoding on the quantized feature map and quantized hyper-prior information involved in an entropy model by using the entropy model to obtain a compressed bit-stream;
entropy decoding: performing entropy decoding on the compressed bit-stream, and sequentially recovering the quantized hyper-prior information and the quantized feature map by using the entropy model;
inverse quantization: performing inverse quantization on the recovered quantized feature map to obtain a reconstructed feature map of the image;
inverse mapping: performing inverse mapping on the reconstructed feature map through a first synthesis transform neural network to obtain a reconstructed image; and
bit-rate control: adjusting quantization and inverse quantization parameters in an encoding process according to a target bit-rate or target distortion so that a bit-rate of the compressed bit-stream is close to the target bit-rate or a distortion of the reconstructed image is close to the target distortion;
the first decomposition transform neural network comprising:
a down-sampling subnetwork, wherein the down-sampling subnetwork is implemented by a convolutional neural network and is configured to transform an input image into a hidden representation; and
a reversible encoding subnetwork, wherein the reversible encoding subnetwork is configured to transform the hidden representation into the initial feature map; the reversible encoding subnetwork is obtained by hierarchical arrangement of reversible encoding units, the hierarchical arrangement comprises/layers, the ith layer (1≤i≤I) comprises 2i−1 reversible encoding units, each of the reversible encoding units is provided with two outputs, and the two outputs of the reversible encoding unit in the ith layer are used as inputs of two reversible encoding units in the (i+1)th layer; for the unique reversible encoding unit in the first layer, an input thereof is the hidden representation; and for 2I outputs provided in the Ith layer, the initial feature map can be obtained by merging.