In this paper, a technique of partial shape recognition is proposed. After polygonal approximation of input scene, feature elements composed of three arcs and two neighboring vertices are extracted. Using feature elements of input scene and those of models in the database, the input scene is compared with models and it is decided which parts of input scene are occluded and which are not. Some models in the database are selected as candidates if there exits a best matching path between input scene and corresponding model. After reconstructing the occluded parts of input scene by models, moment Fourier descriptor is used to finally recognize input scene. Simulation results show that the proposed method recognizes partially occluded objects well.
And efficient database image storage technique is proposed. In database image storage, only object is important and backgrounds are not necessary. So selective coding scheme using EI(Extension-Interpolation) method is applied. The result of EI method for compressing database image showed better performance than conventional still image compression technique(JPEG).
