| 摘要 |
By using genetic programming to create an image-processing filter, a teaching-parameter acquisition unit acquires a teaching parameter that represents a characteristic shape present in unprocessed input images. A data processing unit creates output images by image-processing the input images using the aforementioned image-processing filter, and a characteristic extraction unit extracts a detection parameter that represents a characteristic shape present in the output images. An autoconfiguration unit evaluates the image-processing filter by calculating the cosine similarity between the teaching parameter and the detection parameter. |
