Object Detection in Natural Backgrounds D Predicted by iscrimination Performance and Models

any models of visual performance predict image discriminability, the visibility of the n m difference between a pair of images. We compared the ability of three image discriminatio odels to predict the detectability of objects embedded in natural backgrounds. The three s models were: a multiple channel Cortex transform model with within-channel masking, a ingle channel contrast sensitivity filter model, and a digital image difference metric. Each e d model used a Minkowski distance metric (generalized vector magnitude) to summate absolut ifferences between the background and object plus background images. For each model, this c summation was implemented with three different exponents: 2, 4 and ∞. In addition, each ombination of model and summation exponent was implemented with and without a simple o contrast gain factor. The model outputs were compared to measures of object detectability btained from 19 observers. Among the models without the contrast gain factor, the multiple o channel model with a summation exponent of 4 performed best, predicting the pattern of bserver d′ s with an RMS error of 2.3 dB. The contrast gain factor improved the predictions t of all three models for all three exponents. With the factor, the best exponent was 4 for all hree models, and their prediction errors were near 1 dB. These results demonstrate that. K image discrimination models can predict the relative detectability of objects in natural scenes