Axiomatic approach to computational attention

作者：Garcia J A; Rodriguez Sanchez Rosa; Fdez Valdivia J^*

来源：Pattern Recognition, 2010, 43(4): 1618-1630.

DOI：10.1016/j.patcog.2009.09.027

• Computational attention
• Axiomatic models
• Visual target distinctness
• Military

Here we describe, in terms of a decision problem, any situation in which a computational system will be forced to allocate attention at any time to one spatial location to improve the reconstruction fidelity on a neighborhood of the chosen point. The result is a rational model of computational attention in which a multi-bitrate attention map will provide us with the attention score for each spatial location at high and low quality versions of the image reconstruction. At any time a rational system should choose, even though without any outside knowledge, among alternative spatial locations in such a way as to avoid certain forms of behavioral inconsistency. We compare the performance between a rational approach of computational attention and various models for predicting visual target distinctness, using scenes that represent military vehicles in complex rural backgrounds.

• 出版日期2010-4