In this paper, we consider the problem of image preprocessing for further pattern recognition through the use of a multiagent neurocognitive architecture. The solution to this problem is achieved using the breadth-first search (BFS) method. The article presents algorithmic descriptions of the segmentation method and the image processing method in a multiagent neurocognitive architecture. Experiments were carried out on object recognition in a segmented image based on a multi-agent neurocognitive architecture.