Visual signal detection in structured backgrounds. III. Calculation of figures of merit for model observers in statistically nonstationary backgrounds

Models of human visual detection have been successfully used in computer-ge nerated noise. For these backgrounds, which are generally statistically sta tionary, model performance can be readily calculated by computing the index of detectability d' from the noise power spectrum, the signal profile, and the model template. However, model observers are ultimately needed in more real backgrounds, which may be statistically nonstationary. We investigate d different methods to calculate figures of merit for model observers in re al backgrounds based on different assumptions about image stationarity. We computed performance of the nonpre whitening matched-filter observer with a n eye filter on mammography and coronary angiography for an additive or a m ultiplicative signal. Performance was measured either by applying the model template to the images or by computing closed-form expressions with variou s assumptions about image stationarity. Results show first that the structu red backgrounds investigated cannot be considered stationary. Second, tradi tional closed-form expressions of detectability calculated from the noise p ower spectra with the assumption of background stationarity lead to erroneo us estimates of model performance. Third, the most accurate way of measurin g model performances is by directly applying the model template on the imag es or by computing a closed-form expression that does not assume image stat ionarity. (C) 2000 Optical Society of America. OCIS codes: 330.7310, 330.18 80, 110.3000, 330.4060.[S0740-3232(00)02301-2].