| 摘要 |
A method and apparatus for finding an edge contour in an image of an object is provided that is robust against object/background misclassification due to non-uniform illumination across an image, while also being computationally efficient, and avoiding the need to select a classification threshold. The invention can be used to partition an image of a scene into object regions and background regions, or foreground regions and background regions, using edge contours found in the image. The invention is particularly useful for analysis of images of back-lit objects. The edge contour is progressively formed by finding a sequence of one-dimensional edge positions, each one-dimensional edge position being determined by processing a set of pixels arranged along at least one imaginary line.
|
