Determination of optimal top-down gains for specific searching tasks

Finding optimal top-down feature gains plays a key role in modeling task-driven visual attention mechanisms. Some studies suggest that the ratio of the mean salience of the target to the distractors can be used to determine the weights for the feature maps during the searching process, but this works well only if the salience distribution in the feature map is uniform, which is seldom seen in natural scenes. Here, we derive a new optimal feature gain modulation strategy to maximize the relative salience of the target, in which the top-down weight on a feature map depends on its stimulation intensity ratio (SIR) between the target and the distractors. The stimulation intensity is determined by two factors, i.e., cumulative summation of salience (CSS) and the mean activity coefficient (MAC). Testing on synthetic scenes shows that our model may provide accurate assessment of the contribution of the feature maps in computing the saliency map for a given task.