Estimation of Shape Visibility by Using a Low-pass Filtering Method

Abstract

This paper concerned a low-pass filtering method for evaluating the visibility of shapes. With this method, observers discriminate shapes while their images are filtered in gradually lower spatial frequencies. We estimate the visibility by measuring discrimination thresholds in cut-off frequencies. The method has been validated in practical situations where visual information carried by high spatial frequencies is degraded. Such situations are surprisingly common under unfavor-able conditions, including peripheral vision, mesopic or scotopic vision, isoluminance, image motion, and also in the eyes elderly persons and clinically impaired eyes.
In this experiment, the numerals from 14 font families were estimated by the lowpass filtering method. Eight numeral images, extracted from one of each font family, were rotated clockwise by 0, 90, 180, or 270 degees, and sequentially presented on a CRT display in random order for 200 ms each. The image contrast was about 0.99. Their positions were randomly distributed around the fixation point. The letters were low-pass filtered by a ground glass placed between an observer and the display surface. The cut-off frequency was defined at 1/e of the image peak contrast. The observer was asked to judge whether the target letter, which was presented at the beginning of each session, was included in the stimulus sequence consisting of eight numerals. Threshold cut-off frequency was obtained by a simple staircase method.
Obtained visibility estimates agreed well among observers, and a series of working hypotheses for shapes of more legible letters were suggested.