| 摘要 |
PROBLEM TO BE SOLVED: To solve such a problem that a high-density disparity value image to be derived by a ranging method using an SGM (Semi-Global Matching) can derive a disparity value even in a case of an object with weak texture, however, pixels obtained by capturing the object with weak texture may be sometimes affected by pixels obtained by capturing a distant object with strong texture and reduce disparity accuracy.SOLUTION: A cost synthesizer 320 synthesizes a first weight (e.g., 1) of cost values of a first reference area in a reference image, with a second weight (e.g., 0.9) smaller than the first weight in the cost values of a second reference area in the vicinity of the first reference area in the reference image. When deriving a route cost value Lr' in a pixel at a p location, this constitution can, even when there is a presence of an extremely low cost value C, regardless of being distant from the pixel at the p location, make it less susceptible to be affected by the presence thereof. |
