Spatial integration of band-pass filtered patterns in noise

Contrast energy thresholds were measured for noisy band-pass filtered Sloan letters, K, H, O, and a plus symbol (+) at various centre frequencies and bandwidths. For all patterns the efficiency and energy sensitivity of detection increased with filter bandwidth but decreased with increasing filter centre frequency. Hence, neither parameter alone can explain the changes in efficiency and energy sensitivity. Irrespective of centre frequency and bandwidth, efficiency and energy sensitivity were found to decrease as a single power function of image complexity (Z) defined as the product of the square of the median spatial frequency (fm) of the image energy spectrum and image area (A95) comprising 95% of total contrast energy of the stimulus. The product Z = fm2A95 describes image complexity in the sense that when f2m or A95 increase the number of details increase, too. The decrease of the efficiency of detection with increasing image complexity suggests that the proportion of the integrated image area decreases as the complexity of an image increases.