To detect objects whose properties and constitutions of colors which the object of a reference picture originally has among unknown pictures irrespective of an illumination condition at the time of taking the picture are similar.
In a learning part 10, the reference picture of the object is generated and color feature quantity in the whole reference picture is obtained from a feature quantity calculation part 40. It is transferred to a detection part 20 and the detection part 20 obtains color feature quantity at every local area in the unknown picture from the feature quantity calculation part 40. Color feature quantity is sequentially compared with the color feature quantity of the reference picture and the presence of the object and the position are detected. Then, a detection result is outputted to an output part 30. The feature quantity calculation part 40 calculates the histogram of the ratio of the color value of the picture in the local area and it is set to be color feature quantity.
BUI BUI BINODD
MURASE HIROSHI
