| 摘要 |
PURPOSE: To provide an object recognizing method which can smoothly take out the area of an object which has irregularity in lightness. CONSTITUTION: A difference processing between an image which does not include the object and an image which includes the object, is performed to take out only the part of the object, and pixels included in the area of the object are extracted by multistage binarization. The multistage binarization processing uses a threshold value which is close to the maximum value of lightness of pixels of the object and extracts pixels which are brighter than the threshold value. Then, the threshold value is lowered and pixels which are brighter than the threshold value are added to the area. At this time, only when pixels extracted newly with a new threshold value are connected to pixels that are already determined as pixels constituting the area of the object, they are registered as pixels which constitute the area of the object. Thus, attention to connectivity is paid to update pixels included in the area of the object while the threshold value is varied, so the area of the object can smoothly be found. |
