JPS647787A - Intellectual coding system for image signal - Google Patents

Abstract

PURPOSE:To transmit or accumulate image signals efficiently by generating a feature image to input image signals, extracting the position of an image constituting element and the information of a shape based on the generated feature image, and simply expressing the image with less parameters. CONSTITUTION:A feature image generation part 1 generates a binary image (feature image) for, for example, a section of mouth by using a threshold. A shape information detection part 4 selects a feature point Pi, and separates pairs of coordinates into positional information and shape information. A positional information detection part 2 detects the position of a face in a picture based on the feature image 1, and further detects the approximate positions of elements such as a mouth, an eye, etc., and a shape processing part 4 processes the data. An expression part 3 expresses simply the information by using the numeral as it is or quantizing, variable length coding the numeral. An expression part 5 expresses the information from the detection part 4 by using contour lines or parameter groups, and the parameters are further simply expressed by using the numerals as they are or quantizing, variable length coding the numerals. By such a constitution, particularly a moving image can be expressed/transmitted with less information while maintaining coherency in the variation in the position and shape of the constituent elements.