사진 구도 최적화를 위한 미학적 척도 기반 자동 분석 기술

Abstract

In this paper, a technique based on the measure of aesthetic for optimal photo composition guidance is proposed. Usually, many kinds of photo composition are existed. In our research, we only focus on three of them. There are trichotomy composition, vertical & horizontal composition and centralized composition. For the detection of major object, three kinds of photo images are used. The first one is the photo image which can observe the region of sky. The second one is the close-up photo and telephoto image while the last one is the portraits photo. For the object detection in the first kind of photo image, the standard deviation filter is adopted first which can detect the skyline. Then, the region of sky can be removed and the major object in the photo image can be detected by using image processing method like edge detection and morphology operation. Since the close-up photo and telephoto image always focus on the major object, the major object will present a distinct outline. Therefore, the major object in the close-up photo and telephoto image can be detected by morphology operation after getting the edge information by the method of homogeneity operator. For the major object detection in the portrait photo, the Kinect is used because Kinect can find the human skeletal with the Microsoft Kinect sdk. The major object detected in the different kinds of photo images can be guided appropriately by using three of the pre-defined composition. Compared with other optimal composition guidance technology, our method has reduced the complexity and can obtain better results.