| 摘要 |
Methods and apparatus are disclosed for identifying a particular region in an image based on a user specification and information in the image, such as color and texture information. Initially, the user specifies a region using a single closed contour. The user input specifies the region of interest in the image. The present invention thereafter identifies a region that is (i) near the user input; (ii) about the same size as the initial region; (iii) the most likely boundary based on proximity to the initial boundary, region dissimilarity across the boundary, and similarity in region change along the boundary; and (iv) includes all points of homogeneous sub-regions that are initially mostly included. The present invention identifies the region accurately segmented and appropriately measured. By using local color information, edge information, and the user input, the present invention is capable of segmenting extremely complex regions including regions with highly irregular contours and regions composed of several visually different sub-regions or textures, or both.
|
