DeepFGS: Fine-Grained Scalable Coding for Learned Image Compression

TL;DR

This paper proposes a learned fine-grained scalable image compression framework, namely DeepFGS, which introduces 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.

Abstract

Scalable coding, which can adapt to channel bandwidth variation, performs well in today's complex network environment. However, most existing scalable compression methods face two challenges: reduced compression performance and insufficient scalability. To overcome the above problems, this paper proposes a learned fine-grained scalable image compression framework, namely DeepFGS. 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 one-pass encoding. For entropy coding, we design a mutual entropy model to fully explore the correlation between the basic and scalable features. In addition, we reuse the decoder to reduce the parameters and computational complexity. Experiments demonstrate that our proposed DeepFGS outperforms previous learning-based scalable image compression models and traditional scalable image codecs in both PSNR and MS-SSIM metrics.

Figures (8)

• Figure 1: Illustration of fine-grained scalable Coding of the proposed DeepFGS.
• Figure 2: The architecture of our fine-grained scalable coding framework DeepFGS. Q denotes quantization. AE and AD are arithmetic encoder and decoder.
• Figure 3: Visualization of the entropy and the reconstruction quality of latent representation from kodim20. We divide the 192 channels into 24 groups on average, and the figure shows the entropy and the reconstruction quality of each group.
• Figure 4: The architecture of our mutual entropy model for $\hat{y}_s$.
• Figure 5: Observation of the process of feature fusion. Basic features, fusion features, and difference features refer to: $FFM(\hat{y}_b)$, $FFM(\hat{y}_b || \hat{y}_s)$ and $FFM(\hat{y}_b || \hat{y}_s) - FFM(\hat{y}_b)$. We search for the feature map with the largest activation value among the features and visualize the energy of each channel ($maximum-minimum$).