How lossy compression works
Lossy compression algorithms (such as JPEG and lossy WebP) exploit human visual perception limitations (psychovisual redundancy). The human eye is far more sensitive to variations in luminance (brightness) than chrominance (color).
Codecs convert RGB pixel data into YCbCr color spaces, downsample color channels (4:2:0 subsampling), and perform Discrete Cosine Transforms (DCT) to discard high-frequency visual details that human eyes cannot distinguish.
How lossless compression preserves every pixel
Lossless compression algorithms (such as PNG and lossless WebP) do not discard a single bit of visual information. Instead, they use spatial prediction filters (predicting neighbor pixel colors) and entropy coding algorithms (DEFLATE / LZ77 / Huffman encoding) to represent identical data in fewer bytes.
- Use lossy compression for photographs, hero backgrounds, and complex textures.
- Use lossless compression for logos, typography, screenshots, vector graphics, and transparency.
- Avoid re-saving lossy files repeatedly to prevent generational compression artifacts.