Notes

Chapter 10: Processes of Perception and Analysis

Section 7: Visual Perception


Scale invariance [in vision]

In a first approximation our recognition of objects does not seem to be much affected by overall size or overall light level. For light level—as with color constancy—this is presumably achieved by responding only to differences between levels at different positions. Probably the same effect contributes to scale invariance by emphasizing only edges and corners. And if one is looking at objects like letters, it helps that one has learned them at many different sizes. But also similar cells most likely receive inputs from regions with a range of different sizes on the retina—making even unfamiliar textures seem the same over at least a certain range of scales. When viewed at a normal reading distance of 12 inches each square in the picture on page 578 covers a region about 5 cells across on the retina. With good lighting and good eyesight the textures in the picture can still be distinguished at a distance of 5 feet, where each square covers only one cell. But if the picture is enlarged by a factor of 3 or more then at normal reading distance it can become difficult to distinguish the textures—perhaps because the squares cover regions larger than the templates used at the lowest levels in our visual system.


From Stephen Wolfram: A New Kind of Science [citation]