Data Structures

The large memory requirements associated with storing pictorial data are well known to anyone who has worked with them. For example, storing an ordinary frame of television requires at least 512×512 bytes, if we use three bits for two of the primary colors and two for the third. A black and white passport photograph requires at least a 64×64 matrix with six bits per element, well above the size of a record containing whatever other information is in a passport. (A page of single-spaced typewritten text requires about 3000 bytes.) Problems of storage, search, retrieval, transmission, etc. are particularly difficult whenever pictorial data are encountered. These difficulties are somewhat counterintuitive because humans often find it easier to deal with pictures than with text. It is far easier for us to remember the face of a new acquaintance than a page of typewritten text. The difficulty of matching such human performance on a computer can be appreciated by pointing out that, at least for some people, the recollection of the face is better when it belongs to a member of the opposite sex and that the text is remembered better if it is a piece of prose than if it is a list of names. Therefore, any data compaction techniques that depend only on signal processing are not likely to reduce the data volume to a size compatible with our intuitive expectations.