Digital Image Aesthetic Composition Optimization Based on Perspective Tilt Correction
Hang Shao, Yongxiong Wang, Derui Ding, Chaoli Wang · 2020
In the field of the digital image intelligent aesthetic, many efforts have been made in automatic aesthetic assessment and composition optimization. Comparatively, few studies focus on image tilt correction. In this paper, we propose a novel method for automatically correct image tilt and optimize image visual balance. In our method, the perspective transformation based on Cartesian coordinates is innovatively used to improve the visual balance of the flat image. Meanwhile, in feature extraction, learning-based algorithm is employed as the backbone. In order to further boost the robustness of our method, a novel line segment clustering detection algorithm (LSCD) is proposed to detect line features. Then, the line features are used to calculate the tilt compensation angle. The LSCD algorithm can effectively make up for the defects of the traditional Hough transform line detection algorithm. We perform multiple experiments using images to qualitatively and quantitative verify the scientificity and validity of the proposed method. Experimental results demonstrate that the optimization performance of our method is significantly better than the state-of-the-art straight line-based and affine transformation-based correction algorithm.