Substitutional Neural Image Compression

We describe Substitutional Neural Image Compression (SNIC), a general approach for enhancing any neural image compression model, that requires no data or additional tuning of the trained model. It boosts compression performance toward a flexible distortion metric and enables bit-rate control using a single model instance. The key idea is to replace the image to be compressed with a substitutional one that outperforms the original one in a desired way. Finding such a substitute is inherently difficult for conventional codecs, yet surprisingly favorable for neural compression models thanks to their fully differentiable structures. With gradients of a particular loss backpropogated to the input, a desired substitute can be efficiently crafted iteratively. We demonstrate the effectiveness of SNIC, when combined with various neural compression models and target metrics, in improving compression quality and performing bit-rate control measured by ratedistortion curves. Empirical results of control precision and generation speed are also discussed.

[1]  Li Fei-Fei,et al.  ImageNet: A large-scale hierarchical image database , 2009, CVPR.

[2]  Yao Wang,et al.  End-to-End Learnt Image Compression via Non-Local Attention Optimization and Improved Context Modeling , 2019, IEEE Transactions on Image Processing.

[3]  Jungwon Lee,et al.  Variable Rate Deep Image Compression With a Conditional Autoencoder , 2019, 2019 IEEE/CVF International Conference on Computer Vision (ICCV).

[4]  Taco S. Cohen,et al.  Lossy Compression with Distortion Constrained Optimization , 2020, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[5]  David Minnen,et al.  Variational image compression with a scale hyperprior , 2018, ICLR.

[6]  Touradj Ebrahimi,et al.  The JPEG 2000 still image compression standard , 2001, IEEE Signal Process. Mag..

[7]  Joan Bruna,et al.  Intriguing properties of neural networks , 2013, ICLR.

[8]  Miska M. Hannuksela,et al.  Block-optimized Variable Bit Rate Neural Image Compression , 2018, CVPR Workshops.

[9]  Luc Van Gool,et al.  Generative Adversarial Networks for Extreme Learned Image Compression , 2018, 2019 IEEE/CVF International Conference on Computer Vision (ICCV).

[10]  Aleksander Madry,et al.  Towards Deep Learning Models Resistant to Adversarial Attacks , 2017, ICLR.

[11]  Abdelaziz Djelouah,et al.  Content Adaptive Optimization for Neural Image Compression , 2019, CVPR Workshops.

[12]  Zhou Wang,et al.  Multiscale structural similarity for image quality assessment , 2003, The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003.

[13]  Valero Laparra,et al.  End-to-end Optimized Image Compression , 2016, ICLR.

[14]  David A. Wagner,et al.  Towards Evaluating the Robustness of Neural Networks , 2016, 2017 IEEE Symposium on Security and Privacy (SP).

[15]  Michael W. Marcellin,et al.  JPEG2000 - image compression fundamentals, standards and practice , 2002, The Kluwer International Series in Engineering and Computer Science.

[16]  David Minnen,et al.  Full Resolution Image Compression with Recurrent Neural Networks , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[17]  Long Ye,et al.  CodedVision: Towards Joint Image Understanding and Compression via End-to-End Learning , 2018, PCM.