In this paper, a new method for designing modularized vision systems using a distributed cooperative architectures was introduced. In this method, the target of vision is defined in advance using a set of parameters, the relations between which are determined by primitive function-based vision modules. The parameters themselves are determined simultaneously by running these vision modules under a cooperation mechanism. Using this architecture, three types of vision system were actually implemented and tested. The results experimentallyverified that vision systems designed using the proposed method work successf%lly.