So typically one would think that when creating a binary image through some process, for example thresholding, the object should be represented by white pixels, and the background by black pixels. But as is the case, we more often consider the object as being black on a white background. Why is this? Is it because we are use to reading black text on a white page? Is there some scientific explanation?
However, most studies have shown that dark characters on a light background are superior to light characters on a dark background (when the refresh rate is fairly high). For example, Bauer and Cavonius (1980) found that participants were 26% more accurate in reading text when they read it with dark characters on a light background. Could the same be true for binary images? From the algorithms perspective, it doesn’t care because it doesn’t need to visualize an image. However from a human perspective, dark objects look better on a light background.
Due to certain asymmetries in the human visual processing system, dark text on lighter background (“negative contrast”) also offers higher contrast sensitivity than light on dark (“positive contrast”). The disruptive tendency for white letters to spread out or “bleed” over a black background is called irradiation.
Reference: Bauer, D., & Cavonius, C., R. (1980). Improving the legibility of visual display units through contrast reversal. In E. Grandjean, E. Vigliani (Eds.), Ergonomic Aspects of Visual Display Terminals (pp. 137-142). London: Taylor & Francis