these operations were used to indicate degrees of the information detailing. With respect to computational effi-Human beings perceived real-world objects in a graded manner, with macroviews providing global information about the ciency, the use of operation parameters might be a good objects and microviews providing more detailed information. choice for the index of degree of detailing. However, Consequently, researchers in computer vision have directed using operation parameters does not directly match hutheir efforts toward building graded representations of objects, man intuition.
especially shapes of objects. Most previous approaches em-
To overcome this defect, we devised a graded approach ployed operations such as smoothing, blurring, or grouping to to deriving a shape representation with multiple degrees gradually obtain representations with different degrees of deof detail. Our approach is based on a similarity ratio that tail. Parameters of these operations are used to indicate the measures the similarity between an approximation and an degree of detail. One objection to using operator parameters original contour segment. First, a numerical value repreis that these parameters are difficult for viewers to interpret senting the similarity requirement is stated. Guided by this intuitively. To overcome this objection, we propose using a graded approach guided by a given similarity requirement to value, our approach constructs an approximation of the construct a more concise shape description. Our shape represenshape that satisfies the stated requirement. In our terminoltation matches human intuition about detailing degree and ogy, an approximation of a shape consists of a set of tokens. suits machine requirement for concise description. ยฉ 1996 Aca-A token is a line segment fitted to a partial contour of demic Press, Inc.
a shape. Our approach initially uses only two tokens to approximate the given shape. When an optimal result is obtained, we compare the similarity ratio of these tokens