Geometric Transformation-Embedded Mamba for Learned Video Compression

2026-03-09

AI Summary

This work proposes an efficient end-to-end video compression framework based on a direct transformation strategy, circumventing the need for explicit motion estimation and compensation that complicates traditional learning-based approaches. The method uniquely integrates geometric transformation into the Mamba architecture through a novel Cascaded Mamba Module (CMM) and a locality-refined feed-forward network (LRFFN). It further incorporates a differential convolutional hybrid block and a conditional channel-wise entropy model to enhance coding efficiency. Experimental results demonstrate that the proposed approach significantly improves perceptual quality and temporal consistency at low bitrates, outperforming state-of-the-art methods across multiple evaluation metrics.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Although learned video compression methods have exhibited outstanding performance, most of them typically follow a hybrid coding paradigm that requires explicit motion estimation and compensation, resulting in a complex solution for video compression. In contrast, we introduce a streamlined yet effective video compression framework founded on a direct transform strategy, i.e., nonlinear transform, quantization, and entropy coding. We first develop a cascaded Mamba module (CMM) with different embedded geometric transformations to effectively explore both long-range spatial and temporal dependencies. To improve local spatial representation, we introduce a locality refinement feed-forward network (LRFFN) that incorporates a hybrid convolution block based on difference convolutions. We integrate the proposed CMM and LRFFN into the encoder and decoder of our compression framework. Moreover, we present a conditional channel-wise entropy model that effectively utilizes conditional temporal priors to accurately estimate the probability distributions of current latent features. Extensive experiments demonstrate that our method outperforms state-of-the-art video compression approaches in terms of perceptual quality and temporal consistency under low-bitrate constraints. Our source codes and models will be available at https://github.com/cshw2021/GTEM-LVC.

Problem

• learned video compression
• motion estimation
• motion compensation
• hybrid coding paradigm
• video compression complexity

Innovation

• Mamba
• geometric transformation
• learned video compression
• entropy modeling
• temporal dependency