| 摘要 |
The underlying invention generally relates to the field of object recognition, especially to a contour and surface detection technique (S3) of an object recognition system for digitized images (102) which can advantageously be used for a segmentation of prominent objects, and, more particularly, to an edge detection algorithm (S6) using oriented line segments (edges) to recognize large objects. <??>In the proposed algorithm, an image (102) is segmented into areas of similar homogeneity at a coarse scale, which are then interpreted as surfaces. In this connection, the applied edge detection algorithm (S6) is specially optimized for simultaneously evaluating information from different spatial scales as well as different image features by exploiting (S3) statistical dependencies on their joint appearance. Thereby, the local standard deviation of specific gray levels in the close environment of an observed pixel serves as a measure for local image homogeneity which is used to get an estimate of dominant global object contours. This information is then used to mask (S5a, S5b) the original image (102). Thus, a fine-detailed edge detection (S6) is only applied to those parts of an image (102) where global contours exist. After that, said edges are subject to an orientation detection (S6). Moreover, the algorithm is capable of suppressing noise and small details, thereby contributing to the robustness of object recognition. In contrast to conventional approaches, an automatic parameter adjustment based on the mean gray level of the respective image (102) is proposed. <IMAGE> <IMAGE> |
