Visual attention and target detection in cluttered natural scenes

Rather than attempting to fully interpret visual scenes in a parallel fashi on, biological systems appear to employ a serial strategy by which an atten tional spotlight rapidly selects circumscribed regions in the scene for fur ther analysis. The spatiotemporal deployment of attention has been shown to be controlled by both bottom-up (image-based) and top-down (volitional) cu es. We describe a detailed neuromimetic computer implementation of a bottom -up scheme for the control of visual attention, focusing on the problem of combining information across modalities (orientation, intensity, and color information) in a purely stimulus-driven manner. We have applied this model to a wide range of target detection tasks, using synthetic and natural sti muli. Performance has, however, remained difficult to objectively evaluate on natural scenes, because no objective reference was available for compari son. We present predicted search times for our model on the Search_2 databa se of rural scenes containing a military vehicle. Overall, we found a poor correlation between human and model search times. Further analysis, however , revealed that in 75% of the images, the model appeared to detect the targ et faster than humans (for comparison, we calibrated the model's arbitrary internal time frame such that 2 to 4 image locations were visited per secon d). It seems that this model, which had originally been designed not to fin d small, hidden military vehicles, but rather to find the few most obviousl y conspicuous objects in an image, performed as an efficient target detecto r on the Search_2 dataset, Further developments of the model are finally ex plored, in particular through a more formal treatment of the difficult prob lem of extracting suitable low-level features to be fed into the saliency m ap. (C) 2001 Society of Photo-Optical Instrumentation Engineers.