| 摘要 |
PURPOSE: To provide a feature extraction method which extracts the features of an object out of an input image by the image processing with no disturbance caused by the environmental changes, the image breakage due to the noises, etc., and with reduction of time needed for extraction of the object features. CONSTITUTION: An edge detection filter 901 detects the edge strength at every part of an input image, and an edge calculation means 902 decides the pixels including edges based on the detected edge strength. A 1st projection means 903 performs the projection for the pixels having edges and produces a histogram, and a feature candidate extraction means 904 extracts the feature candidates of an object out of the histogram. A 2nd projection means 905 performs the projection for the areas nearby the feature candidates and produces the second histogram. Then a feature extraction means 906 extracts the object features out of the second histogram to detect the edges of the image inputted from a camera and to extract the features of the object. |
