On the relative complexity of active vs. passive visual search

摘要

Here, this author attempts to tie the concept of active perception to attentive processing in general and to the complexity level analysis of visual search described previously; the aspects of active vision as they have been currently described form a subset of the full spectrum of attentional capabilities. Our approach is motivated by the search requirements of vision tasks and thus we cast the problem as one of search preceding the application of methods for shape-from-X, optical flow, etc., and recognition in general. This perspective permits a dimension of analysis not found in current formulations of the active perception problem, that of computational complexity. This article describes where the active perception paradigm does and does not provide computational benefits along this dimension. A formalization of the search component of active perception is presented in order to accomplish this. The link to attentional mechanisms is through the control of data acquisition and processing by the active process. It should be noted that the analysis performed here applies to the general hypothesize-and-test search strategy, to time-varying scenes as well as to the general problem of integration of successive fixations. Finally, an argument is presented as to why this framework is an extension of the behaviorist approaches to active vision.