Identifying Image Compression Methods Using Vision Transformers

Image compression uses various lossy methods, each creating different distortions. Understanding these distortions is key for optimizing image enhancement and steganography, as they affect quality and detectability. Detection of compression algorithms is essential especially when file extensions are altered. This study presents a method for identifying compression algorithms - JPEG, AVIF, HEIC, and WebP, using a fine-tuned vision transformer model. The model achieved 98% validation and 99% test accuracy and shows premising results.