(Discrete Cosine Transform, DCT)
The requirements to an ideal format for image compression can be outlined as follows:
The image information should be differentiated according to their meaning for the qualitative impression.
The internal structure should offer the opportunity to decide, to what extent the omission of information deteriorates the image quality.
Only relevant information should require data capacity. If a section does not contain information like monochromatic areas, this should be directly reflected within encoded data.
The pixelwise representation used for conventional data formats do not fulfill these requirements. For that reason the original data has to be converted by the help of the so-called Discrete Cosine Transform (DCT). After conversion the data show the following characteristics:
Simple structures are reflected in low values; complex, more detailed structures in high values.
Values resulting from the DCT reflect the geometrical structure of the contents. Details, as horizontal or vertical patterns, not contained in the image will be represented by the value zero.
Monochromatic areas will be described by a single value. All other values become zero.
Normally 8 x 8 pixels of each component will be transformed at a time with the help of the following formula:
< ^ >