Checkerboard Context Model for Efficient Learned Image Compression

He, Dailan; Zheng, Yaoyan; Sun, Baocheng; Wang, Yan; Qin, Hongwei

doi:10.1109/cvpr46437.2021.01453

Cited by 191 publications

(98 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For this reason, though the performance of the first slice degrades (due to hyperprior only), Entroformer counteracts this effect by providing a more powerful context model to promote the performance of the second slice. In the other hand, compared to the CNN-based accelerated method (about 4% performance degradation) [He et al, 2021], our transformer-based method can utilize rich context and achieve a better balance between speed and performance (about 1% performance degradation).…”

Section: Parallel Bidirectional Context Modelmentioning

confidence: 98%

“…In this section, we first propose two ingredients, a diamond relative position encoding (diamond RPE) and a top-k scheme, which are essential for image compression. Then, we extend the checkboard context model [He et al, 2021] to a parallel bidirectional context model.…”

Section: Transformer-based Entropy Modelmentioning

confidence: 99%

“…Following the work of He et al [2021], the latents are split into two slices along the spatial dimension. Figure 4 provides a high-level overview of this architecture.…”

Section: Parallel Bidirectional Context Modelmentioning

confidence: 99%

“…It decodes symbols in a raster-scan order with O(n) serial process that can not be accelerated by modern GPUs. A two-pass parallel context model [He et al, 2021] is introduced for acceleration, which decodes symbols in a particular order to minimize serial processing. However, this parallel context model uses a weak context information, which degrades the compression performance.…”

Section: Introductionmentioning

confidence: 99%

“…In Entroformer, spatial and content based dependencies are jointly taken into account in both hyperprior and context model. (3) The two-pass decoding framework [He et al, 2021] is utilized to speed up the decoding of Entroformer. A bidirectional context with long-range context is introduced instead of the local checkeboard context, which helps counteract the performance degradation.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Entroformer: A Transformer-based Entropy Model for Learned Image Compression

Qian¹,

Ma²,

Sun³

et al. 2022

Preprint

View full text Add to dashboard Cite

One critical component in lossy deep image compression is the entropy model, which predicts the probability distribution of the quantized latent representation in the encoding and decoding modules. Previous works build entropy models upon convolutional neural networks which are inefficient in capturing global dependencies. In this work, we propose a novel transformer-based entropy model, termed Entroformer, to capture long-range dependencies in probability distribution estimation effectively and efficiently. Different from vision transformers in image classification, the Entroformer is highly optimized for image compression, including a top-k self-attention and a diamond relative position encoding. Meanwhile, we further expand this architecture with a parallel bidirectional context model to speed up the decoding process. The experiments show that the Entroformer achieves state-of-the-art performance on image compression while being time-efficient. Code is available at https://github.com/mx54039q/entroformer.

show abstract

Section: Parallel Bidirectional Context Modelmentioning

confidence: 98%

Section: Transformer-based Entropy Modelmentioning

confidence: 99%

“…Following the work of He et al [2021], the latents are split into two slices along the spatial dimension. Figure 4 provides a high-level overview of this architecture.…”

Section: Parallel Bidirectional Context Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Entroformer: A Transformer-based Entropy Model for Learned Image Compression

Qian¹,

Ma²,

Sun³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Expanded Adaptive Scaling Normalization for End to End Image Compression

Shin

Lee

Son

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Contextformer: A Transformer with Spatio-Channel Attention for Context Modeling in Learned Image Compression

Koyuncu

Gao

Boev

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

In this work, we introduce Efficient Contextformer (eContextformer) for context modeling in lossy learned image compression, which is built upon our previous work, Contextformer. The eContextformer combines the recent advancements in efficient transformers and fast context models with the spatio-channel attention mechanism. The proposed model enables content-adaptive exploitation of the spatial and channelwise latent dependencies for a high performance and efficient entropy modeling. By incorporating several innovations, the eContextformer features improved decoding speed, model complexity and rate-distortion performance over previous work. For instance, compared to Contextformer, the eContextformer requires 145x less model complexity, 210x less decoding speed and achieves higher average bit savings on the Kodak, CLIC2020 and Tecnick datasets. Compared to the standard Versatile Video Coding (VVC) Test Model (VTM) 16.2, the proposed model provides up to 17.1% bitrate savings and surpasses various learning-based models.

show abstract

Checkerboard Context Model for Efficient Learned Image Compression

Cited by 191 publications

References 15 publications

Entroformer: A Transformer-based Entropy Model for Learned Image Compression

Entroformer: A Transformer-based Entropy Model for Learned Image Compression

Expanded Adaptive Scaling Normalization for End to End Image Compression

Contextformer: A Transformer with Spatio-Channel Attention for Context Modeling in Learned Image Compression

Contact Info

Product

Resources

About