K均值聚类算法判断单目图像深度层次

Determining the Depth Level of Monocular Image Based on K-Means Clustering Algorithm

针对单目图像深度层次的判断问题提出一种基于K均值聚类算法判断单目图像深度层次的方法。通过图像增强和图像滤波的方式达到优化图像提高对比度,保留细节特征以及滤除图像噪声的目的。根据K均值聚类算法,通过将灰度值相似的像素聚类,进行图像的分割,从而实现深度层次的划分。本次实验选取不同数目的聚类中心进行深度层次的划分并将结果进行比较。实验结果表明,利用K均值聚类的方法判断深度层次在一些几何场景中取得了比较好的成果。

Aiming at the problem of judging the depth levels of monocular images, a method based on the K-means clustering algorithm to judge the depth levels of monocular images is proposed. Through image enhancement and image filtering, the purpose of optimizing the image to improve the contrast, retaining the detailed features and filtering the image noise is achieved. According to the K-means clustering algorithm, the image is segmented by clustering pixels with similar gray values to achieve depth level division. In this experiment, different numbers of clustering centers are selected for indepth division and the results are compared. Experimental results show that the depth level judgment method proposed in this paper has achieved relatively good results in some geometric scenes.