Variable Rate Image Compression with Recurrent Neural Networks

Toderici, George; O’Malley, Sean M.; Hwang, Sung Jin; Vincent, Damien; Minnen, David; Baluja, Shumeet; Covell, Michele; Sukthankar, Rahul

doi:10.48550/arxiv.1511.06085

Cited by 109 publications

(185 citation statements)

References 0 publications

Supporting

Mentioning

184

Contrasting

Order By: Relevance

“…Fig. 4 (b) and (c) show that the SICS is superior or competitive to traditional methods in terms of the reconstruction performance, which is consistent with the opinion of most DNNbased image compression methods [10], [11], [12]. In essence, it is the better quality of reconstruction images that leads to better classification performance.…”

Section: B Comparison Experimentssupporting

confidence: 76%

“…So 𝑑𝑧 𝑘𝑖 𝑗 𝑑𝑒 𝑘𝑖 𝑗 = 1. Toderici et al [12] used a stochastic binarization function as 𝑧 𝑘𝑖 𝑗 = −1 when 𝑒 𝑘𝑖 𝑗 < 0, and 𝑧 𝑘𝑖 𝑗 = 1 otherwise. Here we point out that eithr using a proxy function, or adding an uniform noise, their derivatives of corresponding rounding functions are both 1 in the back propagation.…”

Section: A System Architecturementioning

confidence: 99%

See 1 more Smart Citation

Task-Oriented Semantic Communication with Semantic Reconstruction: An Extended Rate-Distortion Theory Based Scheme

Liu¹,

Tong²,

Sun³

et al. 2022

Preprint

View full text Add to dashboard Cite

Considering the performance of intelligent task during signal exchange can help the communication system to automatically select those semantic parts which are helpful to perform the target task for compression and reconstruction, which can both greatly reduce the redundancy in signal and ensure the performance of the task. The traditional communication system based on rate-distortion theory treats all the information in the signal equally, but ignores their different importance to accomplish the task, which leads to waste of communication resources. In this paper, combined with the information bottleneck method, we present an extended rate-distortion theory which considers both concise representation and semantic distortion. Based on this theory, a task-oriented semantic image communication system is proposed. In order to verify that the proposed system can achieve performance improvement on different intelligent tasks, we apply the basic system trained with classification task to the system with object detection as the target task. The experimental results demonstrate that the proposed method outperforms the traditional and multi-task based communication system in terms of task performance at the same signal compression degree and noise interference degree. Furthermore, it is necessary to consider a compromise between rate-distortion theory and information bottleneck theory by comparing the pure rate-distortion scheme and the pure IB scheme.

show abstract

Section: B Comparison Experimentssupporting

confidence: 76%

Section: A System Architecturementioning

confidence: 99%

Task-Oriented Semantic Communication with Semantic Reconstruction: An Extended Rate-Distortion Theory Based Scheme

Liu¹,

Tong²,

Sun³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The most representative works can be divided into two branches: recurrent neural network (RNN) based models and convolutional neural network (CNN) based models. RNN based models [16,34,35] compress images or residual information from the previous step iteratively, while CNN based models typically transform images into compact latent representations for further entropy coding. Some early works [1,4,33] solve the problem of non-differential quantization and rate estimation.…”

Section: Image Compressionmentioning

confidence: 99%

“…In the early stage, [34] uses RNN structure to compress raw and residual information iteratively. After each iteration, the bitstreams will increase, and the quality of the reconstructed image will be enhanced.…”

Section: Learned Scalable Image Compressionmentioning

confidence: 99%

“…[15] use a layered coding structure to make the bitstream coarse-grained scalable (the bitstream has four independently decodable subsets), but due to the use of multiple codecs, the inferencing process is complicated, and the scalability of the bitstream is limited. The RNNbased method [16,34,35] can generate a coarse-grained scalable bitstream through multiple iterations. However, the encoding-decoding process complexity is too high, and the rate-distortion performance is lower than the CNN-based methods.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

DeepFGS: Fine-Grained Scalable Coding for Learned Image Compression

Ma¹,

Zhai²,

Wang³

2022

Preprint

View full text Add to dashboard Cite

Scalable coding, which can adapt to channel bandwidth variation, performs well in today's complex network environment. However, the existing scalable compression methods face two challenges: reduced compression performance and insufficient scalability. In this paper, we propose the first learned fine-grained scalable image compression model (DeepFGS) to overcome the above two shortcomings. Specifically, we introduce a feature separation backbone to divide the image information into basic and scalable features, then redistribute the features channel by channel through an information rearrangement strategy. In this way, we can generate a continuously scalable bitstream via onepass encoding. In addition, we reuse the decoder to reduce the parameters and computational complexity of DeepFGS. Experiments demonstrate that our DeepFGS outperforms all learning-based scalable image compression models and conventional scalable image codecs in PSNR and MS-SSIM metrics. To the best of our knowledge, our DeepFGS is the first exploration of learned fine-grained scalable coding, which achieves the finest scalability compared with learning-based methods.

show abstract

DeepSIC: Deep Semantic Image Compression

Luo

Yang

Yin

et al. 2018

Neural Information Processing

View full text Add to dashboard Cite

Incorporating semantic information into the codecs during image compression can significantly reduce the repetitive computation of fundamental semantic analysis (such as object recognition) in client-side applications. The same practice also enable the compressed code to carry the image semantic information during storage and transmission. In this paper, we propose a concept called Deep Semantic Image Compression (DeepSIC) and put forward two novel architectures that aim to reconstruct the compressed image and generate corresponding semantic representations at the same time. The first architecture performs semantic analysis in the encoding process by reserving a portion of the bits from the compressed code to store the semantic representations. The second performs semantic analysis in the decoding step with the feature maps that are embedded in the compressed code. In both architectures, the feature maps are shared by the compression and the semantic analytics modules. To validate our approaches, we conduct experiments on the publicly available benchmarking datasets and achieve promising results. We also provide a thorough analysis of the advantages and disadvantages of the proposed technique.

show abstract

Variable Rate Image Compression with Recurrent Neural Networks

Cited by 109 publications

References 0 publications

Task-Oriented Semantic Communication with Semantic Reconstruction: An Extended Rate-Distortion Theory Based Scheme

Task-Oriented Semantic Communication with Semantic Reconstruction: An Extended Rate-Distortion Theory Based Scheme

DeepFGS: Fine-Grained Scalable Coding for Learned Image Compression

DeepSIC: Deep Semantic Image Compression

Contact Info

Product

Resources

About