Summary of Bi-level Spatial and Channel-aware Transformer For Learned Image Compression, by Hamidreza Soltani et al.
Bi-Level Spatial and Channel-aware Transformer for Learned Image Compression
by Hamidreza Soltani, Erfan Ghasemi
First submitted to arxiv on: 7 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Information Theory (cs.IT); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Transformer-based image compression method enhances the transformation stage by considering frequency components within the feature map. It integrates a Hybrid Spatial-Channel Attention Transformer Block (HSCATB) and a Channel-aware Self-Attention (CaSA) module to improve compression performance. Additionally, it introduces a Mixed Local-Global Feed Forward Network (MLGFFN) to enhance information extraction. This framework improves overall compression efficiency by projecting data into a more decorrelated latent space. The proposed method surpasses state-of-the-art LIC methods in rate-distortion performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Image compression is getting better thanks to new learning-based methods. These methods use special kinds of computer networks called Transformers or Convolutional Neural Networks (CNNs). But sometimes they don’t take into account the way images are structured, which can make them less efficient. To fix this, scientists came up with a new Transformer-based image compression method that looks at frequency components in the image. It uses two special blocks: one for handling high and low frequencies, and another to capture information across different parts of the image. This makes it better at compressing images than previous methods. |
Keywords
» Artificial intelligence » Attention » Feature map » Latent space » Self attention » Transformer
