This research harnesses Large Language Models (LLMs) to revolutionize lossless image compression by integrating visual prompts with textual knowledge.

• Introduces a novel approach connecting LLMs with image compression tasks
• Leverages extensive prior knowledge in pretrained language models
• Shows particularly impressive performance with medical images
• Creates more efficient entropy models for better compression

For healthcare professionals, this advancement could significantly improve storage and transmission of diagnostic images while preserving critical details that lossy compression might eliminate.

Large Language Model for Lossless Image Compression with Visual Prompts