Detecting Dominant Vanishing Points in Natural Scenes with Application to Composition-Sensitive Image Retrieval

摘要

Linear perspective is widely used in landscape photography to create theimpression of depth on a 2D photo. Automated understanding of linearperspective in landscape photography has several real-world applications,including aesthetics assessment, image retrieval, and on-site feedback forphoto composition, yet adequate automated understanding has been elusive. Weaddress this problem by detecting the dominant vanishing point and theassociated line structures in a photo. However, natural landscape scenes posegreat technical challenges because often the inadequate number of strong edgesconverging to the dominant vanishing point is inadequate. To overcome thisdifficulty, we propose a novel vanishing point detection method that exploitsglobal structures in the scene via contour detection. We show that our methodsignificantly outperforms state-of-the-art methods on a public ground truthlandscape image dataset that we have created. Based on the detection results,we further demonstrate how our approach to linear perspective understandingprovides on-site guidance to amateur photographers on their work through anovel viewpoint-specific image retrieval system.