Lossy Image Compression with Compressive Autoencoders
arXiv:1703.00395
Abstract
We propose a new approach to the problem of optimizing autoencoders for lossy image compression. New media formats, changing hardware technology, as well as diverse requirements and content types create a need for compression algorithms which are more flexible than existing codecs. Autoencoders have the potential to address this need, but are difficult to optimize directly due to the inherent non-differentiabilty of the compression loss. We here show that minimal changes to the loss are sufficient to train deep autoencoders competitive with JPEG 2000 and outperforming recently proposed approaches based on RNNs. Our network is furthermore computationally efficient thanks to a sub-pixel architecture, which makes it suitable for high-resolution images. This is in contrast to previous work on autoencoders for compression using coarser approximations, shallower architectures, computationally expensive methods, or focusing on small images.
References in corpus (6)
Cited by in corpus (84)
- Image and Video Compression with Neural Networks: A Review
- Soft-to-Hard Vector Quantization for End-to-End Learning Compressible Representations
- CompressAI: a PyTorch library and evaluation platform for end-to-end compression research
- Video Compression With Rate-Distortion Autoencoders
- Causal Contextual Prediction for Learned Image Compression
- Deep Learning-Based Video Coding: A Review and A Case Study
- Low Bit-Rate Speech Coding with VQ-VAE and a WaveNet Decoder
- Deep Contextual Video Compression
- Real-Time Adaptive Image Compression
- Towards Image Understanding from Deep Compression without Decoding
- Review: Deep Learning in Electron Microscopy
- Quaternion Convolutional Neural Networks for Heterogeneous Image Processing
- Universal Joint Feature Extraction for P300 EEG Classification using Multi-task Autoencoder
- Structure-preserving neural networks
- Deep Learning-based Implicit CSI Feedback in Massive MIMO
- Denoising and Regularization via Exploiting the Structural Bias of Convolutional Generators
- Quantum Hamiltonian-Based Models and the Variational Quantum Thermalizer Algorithm
- Theory and Experiments on Vector Quantized Autoencoders
- Fixing a Broken ELBO
- Learning to Inpaint for Image Compression
- Machine Learning Techniques to Construct Patched Analog Ensembles for Data Assimilation
- Improved Lossy Image Compression with Priming and Spatially Adaptive Bit Rates for Recurrent Networks
- BlockCNN: A Deep Network for Artifact Removal and Image Compression
- Lossy Image Compression with Normalizing Flows
- Generalized Octave Convolutions for Learned Multi-Frequency Image Compression
- Deep Perceptual Compression
- Video Compression through Image Interpolation
- Learning Convolutional Networks for Content-weighted Image Compression
- A Semantic-based Medical Image Fusion Approach
- Deep Generative Models for Distribution-Preserving Lossy Compression
- End-to-End Rate-Distortion Optimization for Bi-Directional Learned Video Compression
- Distributed Learning and Inference with Compressed Images
- Image compression optimized for 3D reconstruction by utilizing deep neural networks
- LOGAN: Membership Inference Attacks Against Generative Models
- BINet: a binary inpainting network for deep patch-based image compression
- A Cross Channel Context Model for Latents in Deep Image Compression
- Regularizing linear inverse problems with convolutional neural networks
- GAN- vs. JPEG2000 Image Compression for Distributed Automotive Perception: Higher Peak SNR Does Not Mean Better Semantic Segmentation
- LatentPoison - Adversarial Attacks On The Latent Space
- Scalable Model Compression by Entropy Penalized Reparameterization
- Gated Context Model with Embedded Priors for Deep Image Compression
- DSSLIC: Deep Semantic Segmentation-based Layered Image Compression
- Multi-bin Trainable Linear Unit for Fast Image Restoration Networks
- Adversarial Defense of Image Classification Using a Variational Auto-Encoder
- Adversarial Video Compression Guided by Soft Edge Detection
- Autoencoder based image compression: can the learning be quantization independent?
- Fully Convolutional Model for Variable Bit Length and Lossy High Density Compression of Mammograms
- DVC: An End-to-end Deep Video Compression Framework
- Slimmable Compressive Autoencoders for Practical Neural Image Compression
- CAE-ADMM: Implicit Bitrate Optimization via ADMM-based Pruning in Compressive Autoencoders
- Toward Joint Image Generation and Compression using Generative Adversarial Networks
- Defending Against Image Corruptions Through Adversarial Augmentations
- Infomax Neural Joint Source-Channel Coding via Adversarial Bit Flip
- End-to-end Optimized Video Compression with MV-Residual Prediction
- CompressNet: Generative Compression at Extremely Low Bitrates
- Learning to compress and search visual data in large-scale systems
- Attention Based Image Compression Post-Processing Convolutional Neural Network
- SZ3: A Modular Framework for Composing Prediction-Based Error-Bounded Lossy Compressors
- Recent Advances in Convolutional Neural Network Acceleration
- Learning to Structure an Image with Few Colors
- Learned Variable-Rate Image Compression with Residual Divisive Normalization
- Neural Image Compression and Explanation
- Neural Multi-scale Image Compression
- DeepSIC: Deep Semantic Image Compression
- Attention-guided Image Compression by Deep Reconstruction of Compressive Sensed Saliency Skeleton
- Stochastic Bottleneck: Rateless Auto-Encoder for Flexible Dimensionality Reduction
- Exploring Autoencoder-based Error-bounded Compression for Scientific Data
- Convolutional Autoencoders for Lossy Light Field Compression
- Neural Communication Systems with Bandwidth-limited Channel
- Saliency Driven Perceptual Image Compression
- Deep Learning-based Image Compression with Trellis Coded Quantization
- Layered Image Compression using Scalable Auto-encoder
- A Deep Image Compression Framework for Face Recognition
- Deep Clustering of Compressed Variational Embeddings
- A Foreground-background Parallel Compression with Residual Encoding for Surveillance Video
- Bit-level Optimized Neural Network for Multi-antenna Channel Quantization
- Variable Rate Image Compression Method with Dead-zone Quantizer
- Source-Aware Neural Speech Coding for Noisy Speech Compression
- Learned Multi-Resolution Variable-Rate Image Compression with Octave-based Residual Blocks
- Object Detection-Based Variable Quantization Processing
- Lossy Medical Image Compression using Residual Learning-based Dual Autoencoder Model
- Unified Signal Compression Using a GAN with Iterative Latent Representation Optimization
- Customized OCT images compression scheme with deep neural network
- Supervised Encoding for Discrete Representation Learning