The visual analysis of non rigid motion is becoming a major area in computer vision. A classification of non rigid motion performed by objects on various degrees of non rigidity is elucidated in this paper according to the curvature changes of the objects

indicating that an 8 class scheme can cover all the objects from rigidity to high non rigidity. A large number of existing approaches in this area are surveyed and divided into two major categories: feature based methods and model based methods

and the advantages and disadvantages of the two classes are discussed respectively. The last section of this paper expatiates on the difficulties confronted by non rigid motion analysis and expects its possible future directions. Non rigid motion analysis is a burgeoning research area but still in its early development. The available body of research has begun to make inroads into only a few of the many difficult problems. Most of the extant models are far from perfect and people are far from a general solution to this problem. Some related research fields

such as speech recognition and computer graphics

may benefit this field.