PNG compression can preserve every original pixel value.

Lossless means the compression process is reversible: decoding recovers the encoded image samples exactly. Repeatedly compressing and decompressing unchanged PNG image data does not gradually introduce the block patterns or blurred edges associated with ordinary lossy JPEG encoding.

PNG is especially useful for screenshots, diagrams, interface graphics, and text because sharp boundaries and repeated colors often compress effectively. Photographs can also be stored exactly, but their complex detail commonly produces larger files than a visually acceptable JPEG.

Transparency is part of the format rather than a workaround requiring a fixed background. PNG can represent partially transparent pixels, helping software composite soft edges, shadows, and other graphics over different backgrounds without baking those backgrounds into the image.

PNG emerged during the 1990s as an alternative to GIF, amid concerns about patents affecting the LZW compression used by GIF. Its initial specification appeared in 1996. The project combined an openly documented format with features useful for exchanging still images across different computer systems.

The name stands for Portable Network Graphics. Unlike a format designed mainly to reduce photographic file sizes at an acceptable visual cost, PNG was built around lossless image storage. It supports grayscale, truecolor, and indexed images, with different capabilities for color depth and transparency depending on the image type.

Its compression pipeline first filters scanline data to expose patterns that compress well. Those filters do not permanently remove visual detail: a decoder can reverse them. The filtered bytes then undergo DEFLATE compression, which exploits repetition and statistical patterns. Different filter choices and compression settings can produce different file sizes while preserving the same decoded image samples.

The important boundary is what enters the encoder. If an editor first reduces an image from higher precision to lower precision, changes its color space, or discards colors, saving the result as PNG cannot restore what the editor already removed. PNG compression is lossless, but a complete editing or export workflow is not automatically lossless.

Exact stored samples also do not guarantee identical appearance everywhere. Color management, display calibration, scaling, and application behavior can change how those samples look on screen. Similarly, a program may preserve the visible image while stripping metadata when it resaves a file. Pixel preservation and preservation of every original file byte are different promises.

PNG, JPEG, and GIF make different promises about image data. Comparing them requires identifying the stage at which information may change. A lossless encoder preserves its input samples, but it cannot restore detail already removed by an earlier resize, palette conversion, or lossy export.