Object detection in natural backgrounds predicted by discrimination performance and models

Abstract

Many models of visual performance predict image discriminability, the visibility of the difference between a pair of images. We compared the ability of three image discrimination models to predict the detectability of objects embedded in natural backgrounds. The three models were: a multiple channel Cortex transform model with within-channel masking; a single channel contrast sensitivity filter model; and a digital image difference metric. Each model used a Minkowski distance metric (generalized vector magnitude) to summate absolute differences between the background and object plus background images. For each model, this summation was implemented with three different exponents: 2, 4 and infinity. In addition, each combination of model and summation exponent was implemented with and without a simple contrast gain factor. The model outputs were compared to measures of object detectability obtained from 19 observers. Among the models without the contrast gain factor, the multiple channel model with a summation exponent of 4 performed best, predicting the pattern of observer d's with an RMS error of 2.3 dB. The contrast gain factor improved the predictions of all three models for all three exponents. With the factor, the best exponent was 4 for all three models, and their prediction errors were near 1 dB. These results demonstrate that image discrimination models can predict the relative detectability of objects in natural scenes.

Detecting changes between real-world objects using spatiochromatic filters

“…Filter-based models have been enormously successful in describing early visual processing, and it is not my goal at this time to apply each of these models to a change detection task or to advocate one of these models over another (Bell & Sejnowski, 1997;Graham, 1992;Lades et al, 1993;Malik & Perona, 1990;Rohaly, Ahumada, & Watson, 1997;see Landy & Movshon, 1991, for a review). 7 Rather, my goal is to distill from these models a simplified and easily implemented approach tailored to the demands of a change detection task.…”

Constrained sampling experiments reveal principles of detection in natural scenes

“…Although there have been a number of studies of detection in natural backgrounds (16)(17)(18)(19)(20)(21)(22)(23), they have not directly addressed these questions, and have either tested only a small number of natural stimuli (16,17,19,20), tested natural stimuli with altered statistical properties (21,22), or used experimental paradigms not representative of natural detection tasks (16,(18)(19)(20)23). These latter studies are not as representative of natural tasks, because observers were allowed to directly compare the same image with and without the added target, an advantage that is not normally available under real-world conditions.…”

The RGFF representational model: a system for the automatically learned partitioning of "visual patterns" in digital images

“…It is therefore of great practical value to have computational visual differences or distinctness measures which can be applied to evaluate image displays, (virtual) scene generators, image compression methods, image reproduction methods, camouflage measures, and traffic safety devices [9], [25]. Rohaly et al [26] recently showed that image discrimination models that quantify the visibility of the differences between a pair of images can predict the visual distinctness of objects in natural backgrounds. Often implicit in the interpretation of visual search tasks is the assumption that the detection of targets is determined by the feature-coding properties of low-level visual processing [9].…”