ALLONLINE STORIES · VOL. III · TRANSPARENCY EXPLAINED
PNG
Transparency became information. How an open format turned alpha, scanline filters and lossless compression into one of the most reliable ways to transport digital graphics.
PNG does not guess what you can stop seeing. It preserves what you exported.
Born in 1996 as an open alternative for web graphics, PNG found a precise role: interfaces, screenshots, illustrations and assets where edges, colour and transparency need to survive intact.
A PNG is a sequence of chunks with very clear rules.
The signature identifies the file. IHDR defines its essential properties. One or more IDAT chunks contain compressed image data. IEND closes the structure. Palettes, colour profiles, text and other auxiliary pieces can live between them.
The eight-byte signature identifies the file before any chunk is interpreted.
Transparency is not an effect. It is per-pixel information.
In truecolor with alpha, every pixel can store red, green, blue and an independent alpha channel. Separating them explains why PNG can preserve soft edges without painting a background into the image.
4 active channels · full colour + transparency
The same pixel can live on any background.
Alpha describes coverage or opacity. The browser blends the pixel colour with whatever is behind it, so a semi-transparent edge can look natural on white, black, colour or an image.
Before DEFLATE, PNG tries to make each scanline more predictable.
Each row can use None, Sub, Up, Average or Paeth. The filter removes no information: it temporarily transforms values so patterns and small differences appear and compress better afterwards.
Lossless means rebuilding exactly the same values.
After filtering, PNG uses DEFLATE: repeated-sequence matching combined with statistical coding. The goal is to describe the same data with fewer bits, not approximate it.
Sometimes 32 colours are smarter than millions.
PNG can store indexed colour through a palette or direct colour per pixel. For icons, pixel art and graphics with few tones, a compact palette can reduce the file substantially without making it lossy.
An interlaced image can reveal itself in seven passes.
Adam7 distributes pixels across seven maps. Early passes provide a sparse structure; each later pass fills the gaps until the complete image is reconstructed. It changes arrival order, not the final result.
PNG also learned to move.
APNG adds multiple frames, duration, blending and disposal to the PNG ecosystem. It is useful for animation with transparency, although web delivery should still compare weight and support with WebP and other modern formats.
A PNG can carry more context than you can see.
Ancillary chunks can describe gamma, chromaticities, ICC profiles, sRGB intent and text. That information can improve consistency between applications, but it also adds bytes and complexity that should be kept deliberately.
PNG wins when every pixel matters. It loses when every byte matters more.
Interfaces, rasterized logos, screenshots, text-heavy graphics and transparency are natural PNG territory. A large photograph usually compresses much better with formats designed for perceptual loss.
Text, icons and hard edges stay crisp because compression does not approximate pixels.
Prioritize correct dimensions and lossless compression.Optimizing PNG means choosing structure before chasing a KB number.
Trim dimensions, decide whether alpha is needed, reduce the palette when appropriate, keep only useful metadata and check whether WebP fits the destination better.
- 01Size to the real display dimensions.
- 02Decide whether you need alpha, truecolor or a palette.
- 03Optimize filters, palette and metadata.
- 04Compare with WebP when the goal is web delivery.
Keep every pixel. Remove only what is unnecessary.
Compress PNG, inspect alpha, extract palettes or compare a WebP output. Every tool runs directly in your browser.
AllOnline Stories · Understanding transparency changes how you export it.