Visual attention is one of the most important issues for a mobile robot to accomplish a given task in complicated environments since the vision sensors bring a huge amount of data. This paper proposes a method of sensor space segmentation for visual attention control that enables efficient observation taking the time needed for observation into account. The efficiency is considered from a viewpoint of not geometrical reconstruction but unique action selection based on information criterion regardless of localization uncertainty. The method is applied to a four legged robot that tries to shoot a ball into the goal. To build a decision tree, a training set is given by the designer, and a kind of off-line learning is performed on the given data set. Discussion on the visual attention control in the method is given and the future issues are shown.