基于单幅图像的自动抠图技术研究

The study of automatic matting on single image

数字抠图是计算机视觉中很重要的领域。通过提取出图像的前景并与新的背景进行合成,在工业中有很多应用。目前,主流抠图方法中人工参与信息是最为重要的先决条件。针对这一问题,本文提出一种新的方法。通过镜面反射原理,获取物体两个不同方向的视觉信息,在现有光流机制的辅助下计算出其景深信息;然后,由算法自动生成trimap,完成自动抠图的整个流程。本文提出方法具有设备简单和效果稳定的优点。最后,通过实例验证了本文的方法的有效性。

Image matting is one of the most important area in computer vision, which extracts objects from the given images and composites it with a new background. There are many applications in image industry. Human interaction is the prior condition in major existed matting approaches. In order to solve this problem, we propose a novel method. Two visual angle information can be received by the application of mirror reflection. As a result, the image depth can be calculated through the existed optical flow mechanism. After that, the trimap can be generated by our algorithm for completing the automatic matting process. The proposed method is stable and easy to do. There are several examples to show the effectiveness and efficiency of our approach.