ݺߣ

ݺߣShare a Scribd company logo

PR-328: End-to-End Optimized Image Compression

?
?
H
Hyeongmin Lee

PR 328?? ??? ICLR 2017? ??? "End-to-End Optimized Image Compression"??? ?????. ??? ??? ?? ???? ?? ?????? ???? ? ?? ??, ? ? ?? ??? ???? ???? ?? ??? ?? ??? ???? ????. ?? ???? ??? JPEG??? ? ? ?????, ? ????? End-to-End Deep Learning? ???? ???? ???? ??? ?????. ? ???? ??? ??? ??? ??? ??? ??? ?? ????? ?? ?????? ??? ????? ??? ??? ???? ???? ????? ????? ???? ???? ???????? :) paper link: https://arxiv.org/abs/1611.01704 youtube link: https://youtu.be/rtuJqQDWmIA

1 of 28
Downloaded 14 times
End-to-End Optimized Image Compression Hyeongmin Lee PR-328 2021.6.27
Entropy Coding
Entropy Coding ? Image Compression 0011010100111...
Entropy Coding ? Entropy Coding A,B,C,D: 4 Letters (need at least 2 bits per letter) Sample Bits A 00 B 10 C 01 D 11 100100001111110101000100010001 (30 bits) https://www.programiz.com/dsa/huffman-coding
Entropy Coding ? Entropy Coding A,B,C,D: 4 Letters (need at least 2 bits per letter) Sample Bits A 11 B 100 C 0 D 101 1000111110110110100110110110 (28 bits) Lower Bound?? Huffman Coding
Entropy Coding ? Entropy ?? = ? ? ???????????????? Sample ???? A 5/15 B 1/15 C 6/15 D 3/15
Image Compression
Image Compression ? Image Compression 0011010100111... Lossless Coding: bmp ? Coding the Integers Itself Lossy Coding: JPEG ? Reducing the Entropy of samples
Image Compression ? JPEG 8x8 cutting Discrete Cosine Transform Quantization Entropy Coding... https://www.whydomath.org/node/wavlets/basicjpg.html
Image Compression ? JPEG High Quality High Entropy High Bitrate Low Quality Low Entropy Low Bitrate
Image Compression ? Bitrate-Distortion Tradeoff
End-to-End Optimized Image Compression [Network Structure]
Network Structure ? Image Compression Pipeline Transform Quantization Entropy Coding Decoding Inverse Transform
Network Structure ? Image Compression Pipeline Transform Quantization Entropy Coding Decoding Inverse Transform Neural Net Neural Net Approximation
Network Structure ? Overall Structure
Network Structure ? Nonlinear Transformation Encoder GDN
Network Structure ? Nonlinear Transformation Decoder IGDN
Network Structure ? Nonlinear Transformation Structure
End-to-End Optimized Image Compression [Optimization]
Optimization ? Loss Function MSE Loss
Optimization ? Entropy of Quantized Vectors
Optimization ? Continuous Relaxation ? ?? = ?? + ??? ~ ??(? ?? ?? , ?? ?? ) ??? ?? ?? = ? ? ???? ?? ?? ? ?? ??, ?? ?? (?? ? ??) ???? = ? ??? ?? ?? ??+ ?? ?? ???? ?? ???? = ????(??)
Optimization ? Loss Functions Piecewise Linear Approximation of ??? ????
Optimization ? Relationship to VAE
Experimental Results
Experimental Results ? Quantitative Evaluation
Experimental Results ? Visual Comparison
Thank You!

More Related Content

PR-328: End-to-End Optimized Image Compression