Dithering isn't always something you should try to avoid.


Confused by a particular term, see our overview below

  • Bit depth - When an image is described as "x bit" with x being some number or other, what's being talked about is the number of colours. In bitplane or raster graphics, each pixel has its colour described by a string of bits, and the more bits there are per pixel the more possible colours there are. The number of colours equals two to the power of the bit depth, so one bit (or "one bitplane") files can have only two colours, two bit can have four colours, three bit has eight colours and so on. The most common depths are 8 bit (256 colours) and 24 bit (16.8 million colours). Bit depths higher than 24 provide a wider colour "gamut", so image manipulation software can pull out otherwise invisible detail out of the image.

  • Bitmapped file - Bitmapped files, also known as raster files, contain graphics information described as pixels, such as photographic images. The image is built up dot by dot; if you zoom in, the pixels get bigger and the image ends up looking like Lego.

  • Compression - Data compression is not a new concept - it's been around in one form or another for decades. If an image format includes data compression, then generally speaking the images will be smaller in size but take more computing power to load, as the computer has to work out what the original data was. In the olden days, compressed formats weren't popular because processors were too slow to display compressed images quickly. This, plus modern advances in compression technology, explains why older image formats tend to be bigger.

  • Dithering - What do you do if you have to display an image with lots of colours on a screen without enough to show it properly? You do dithering. This involves mixing pixels of the colours you have so that the end result looks more like the colour you don't have. It's not as good as having enough colours to show the image properly, but it's better than the "banding" that results from doing a best-match sort of display.

  • Lossy compression - A form of compression in which some data is discarded to allow much smaller file sizes. In image compression, lossy techniques such as those uses in the JFIF format

  • Metafiles - These are files that may contain either bitmapped or vector graphics data.

  • Page Description Languages - PDLs, as they're more often called, are used to describe the layout of a printed page of graphics and text. Two examples are Postscript and HPGL (the Hewlett Packard version). They're used almost exclusively in desktop publishing, most often as the file format sent from the computer to the printer.

  • Palette - The number of colours on screen is not necessarily the same as its palette. The palette is used in the same sense as a painter's palette; it holds all the colours that can be used, from which the ones that actually are used are chosen. A given video card might, for example, allow you to display 256 colours at a time from a palette of 32,768.

  • Pixel - The smallest element of an image, and the unit in which its resolution is expressed. The normal resolution of VGA graphics is 640x480 pixels; such a screen has a total of 307,200 pixels. A low resolution 320x200 image has only 64,000 pixels; a high res 1024x768 image has 786,432.

  • Raster file - See bitmapped file.

  • Vector file - Bitmapped files describe a picture in terms of pixels, while vector files describe it in terms of geometry. A line here, a curve here, this area filled with this colour, and so on. Vector files are much larger for an image of a given detail level, but they can be magnified as much as you like without turning into giant pixels.