| 摘要 |
PROBLEM TO BE SOLVED: To provide a learning type image classifying device which segments a classification object area and reduces a manual work burden such as an instruction. SOLUTION: A classification category contents receiving part 102 receives a classification- desired image that a user inputs and displays to/on an image displaying part 101, and an area segment mode selecting part 103 selects an area corresponding to a classification category. The user selects an image for learning with an image for learning selecting part 104, and an area segment performing part 105 segments a classification category area in a designation mode. A teacher signal attaching part 106 gives a classification category name to the classification category area, a characteristic extracting part 107 extracts characteristics such as the color, shape, etc., of the area, and a classification parameter learning part 108 decides characteristic parameter learning and finishes the learning process. The image is next inputted to a classification object image inputting part 109, a characteristic quantity extracting part extracts the characteristic quantity of each area divided by an area dividing part 110, classifies the characteristic quantity by using a learned classification parameter, and an attaching part 112 gives a category name decided by a category deciding part 111 to the image and finishes the classification process. |
