JPEG - in detail

Definition: The technique - "jay-pegging" - consists of three steps: (i) The Discrete Cosine Transform (DCT): an algorithm similar to that of the Fourier Transform that takes data in blocks measuring 8 pixels x 8 pixels and converts them from the spatial domain to the frequency domain - somewhat similar to presenting a graph of a continuous curve as histograms that plot the frequency of occurance of each value. This step compresses data, loses no detail and identifies data that may be removed. (ii) Secondly, matrix multiplication: the data is re-ordered for quantization: it is here that variable amounts of the data can be discarded through the choice of the quantization coefficient. This is the bit that you are controlling when you choose 'low', 'medium' or 'high' quality or adjust sliders: you must balance your desire for small files with inevitable loss of image quality. (iii) Coding: the results of the last manipulation are finally coded using yet more compression techniques (including a sneaky little zig-zag trick that makes the most of the fact that the last procedure produces data with a lot of zeros in it) but this time with no loss of data. The cumulative effects of three separate steps of compression help to acheive JPEG's very high compression rates: typically files can be reduced by 70% (to 30% original size) with very little visible image degradation; even with 94% reduction, the image is usable for certain purposes e.g. at small sizes or if part of movie. It's thanks to the clout of CCITT (now ITU) and ISO who put their weight behind JPEG (hence 'Joint' in name) that it is an established, clean standard (unlike TIFF that comes in many flavours) which is reliable and widely used. But a new standard, JPEG2000 will supplant its, now dated, technology.

Previous Term: JPEG  Next Term: Judas optics

Type a photography term below to find its definition: