Comparative survey of image compression methods across different pixel formats and bit depths

Abstract Image compression is essential to reduce memory requirements while maintaining a good visual quality of images. The size of raw data of a commonly used RGB 8-bit 4K image would be almost 25 MB, which is too large for efficient streaming, storing and processing. This paper mainly evaluates and compares the state-of-the-art lossy image compression methods. The survey of existing studies is supplemented by an experimental evaluation. The main contribution of this research is the comprehensive comparison taking into account different pixel formats. Modern video compression methods are also used, as they can compress a single image too. Real-life photos are compressed using existing encoders, and visual quality, compression ratio, and encoding speed are evaluated using several metrics. The novel video encoding VVC outperforms other methods in most of the cases. WebP seems to be a viable choice when video encoding is not an option. VVC, AV1, and XVC show very close results with H.265 slightly behind them. The computational complexity of VVC might be problematic when fast processing is necessary, and the other formats might be better options. The paper presents detailed results regarding visual quality, storage requirements, computational time, differences between the compression artifacts and pixel formats, etc.