文摘
We simulate the mechanism of hypercolumns in the V1 cortex of mammals that selectively responds to bar stimuli, and design an orderly-arranged array to extract and represent edges in an image. Based on the neighborhood of units in the array, we construct a graph whose node represents a short line segment of a contour. We search along the routes in that graph and compare them with a shape template for object detection. Organizing segments of contour into a graph greatly upgrades the level of image representation, remarkably reduces the load of combinations, significantly improves the efficiency of object searching, and facilitates the intervening of high-level knowledge.