Every raster image is a grid of pixels — but what exactly does one pixel store? The answer decides how many colours an image can show, whether it can be transparent, how much editing it can survive, and why some conversions quietly change how your image looks. This article unpacks the four properties hiding inside every pixel: bit depth, channels, the alpha channel, and indexed palettes.
The Universal Image Converter exposes all of these directly — you can export to 8-bit or 16-bit, grayscale or RGB or CMYK, with or without alpha — so understanding them turns its options from jargon into deliberate choices.
What a pixel stores
A colour pixel is usually three numbers: how much red, green and blue light it emits. Mix those three and you can make any colour your screen can show — full red + full green = yellow, all three at full = white, all three at zero = black. This is the RGB model, and it exists because that is literally how screens work: every display pixel is a tiny red, green and blue light.
Bit depth: how many levels per channel
Bit depth is how finely each channel is measured. The everyday standard is 8 bits per channel: 256 levels of red × 256 green × 256 blue = 16.7 million colours, far more than the eye can distinguish in a single image. That is why 8-bit is fine for final display.
| Bit depth | Levels / channel | Use |
|---|---|---|
| 1-bit | 2 (black/white) | Fax, line art, masks |
| 8-bit | 256 | Standard photos and graphics |
| 16-bit | 65,536 | Editing masters, scans, medical/scientific |
| 32-bit float | continuous | HDR, rendering, compositing |
Why bother with more than 8-bit if the eye cannot see the difference? Editing headroom. When you push exposure, contrast or colour hard, an 8-bit image can run out of intermediate levels and develop banding — visible steps across a smooth sky. 16-bit has so many spare levels that the same edits stay smooth. The advice: edit in 16-bit, export the final to 8-bit. For very bright highlights and deep shadows beyond what a normal image holds, HDR formats (OpenEXR, Radiance HDR) store light as floating-point values.
Channels: RGB, grayscale, CMYK
RGB is not the only way to describe colour. Different jobs use different channel sets:
- Grayscale — a single channel of brightness, no colour. One-third the data of RGB; perfect for black-and-white photos, scans and masks.
- RGB — red, green, blue. The model for screens and the web.
- CMYK — cyan, magenta, yellow, black: the subtractive inks of print. Printers need CMYK because ink absorbs light rather than emitting it. Converting RGB → CMYK can shift vivid colours, because some bright screen colours simply cannot be made with ink (they are “out of gamut”).
For converting between colour notations within RGB — hex, HSL, and the perceptual spaces — see the colour-converter tools; this article is about how the channels are stored in an image file.
The alpha channel and transparency
Add one more channel to RGB and you get RGBA — the A is alpha, recording how opaque each pixel is, from 0 (fully transparent) to maximum (fully solid). Alpha is what lets an image have a transparent background, or a soft feathered edge that blends onto whatever is behind it.
Crucially, not every format has an alpha channel. PNG, WebP, AVIF, TIFF and GIF (1-bit on/off) can store transparency; JPEG cannot. When you convert a transparent image to a format without alpha, the transparent pixels must be filled with a solid colour and the transparency is lost for good.
Indexed colour and palettes
For images with only a handful of colours, storing a full RGB value per pixel is wasteful. Indexed colour stores a small palette — say 256 chosen colours — and then just a palette index per pixel. A logo with eight colours becomes tiny. This is exactly how GIF and 8-bit PNG work, and why both are great for flat graphics and terrible for photographs: a photo has thousands of subtly different colours that cannot be squeezed into 256 without visible posterisation (flat blotches) or dithering (scattered dots that fake the missing shades).
Conversion gotchas to remember
- To JPEG / BMP / any opaque format → transparency is flattened to a background colour. Choose the fill, or keep PNG/WebP/AVIF.
- RGB → CMYK → very saturated colours may shift to stay printable.
- 16-bit → 8-bit → fine for final output, but do it after heavy editing, not before.
- Photo → indexed (GIF / PNG-8) → expect banding or dithering; reserve it for graphics.
- Colour → grayscale → colour information is discarded and cannot be recovered.
None of these are bugs — they are the unavoidable consequences of what each format can physically store. Knowing them means you choose conversions on purpose instead of being surprised by the result. Put it into practice in the Universal Image Converter, and head back to Image Formats Explained for the big picture.
Frequently asked questions
What is the alpha channel?
Alpha is an extra value stored alongside red, green and blue that records how opaque each pixel is, from fully transparent to fully solid. It is what lets a PNG have a transparent background or a soft, semi-transparent edge. Formats without an alpha channel — like JPEG — cannot store transparency.
What is the difference between 8-bit and 16-bit images?
It refers to bits per channel. 8-bit stores 256 levels per channel (16.7 million colours total) and is fine for final display. 16-bit stores 65,536 levels per channel, giving far more headroom for editing — adjusting exposure or contrast heavily on an 8-bit image can cause visible banding that 16-bit avoids.
Why did my image get a white background after converting to JPEG?
JPEG has no alpha channel, so the transparent areas have to be filled with a solid colour — white by default. The transparency is permanently flattened. To keep transparency, convert to PNG, WebP or AVIF instead.
What is indexed colour?
Instead of storing a full colour for every pixel, an indexed image stores a small palette (say 256 colours) and then a palette index per pixel. It makes graphics with few colours much smaller — this is how GIF and 8-bit PNG work — but is unsuitable for photographs, which have far more than 256 distinct colours.