Various types of noises are unavoidable during the three-dimensional reconstruction of a point cloud. In the existing methods, noise points are usually removed by analyzing and calculating the geometric relation of the point cloud in a three-dimensional space; this involves complicated and time-consuming calculations. This paper proposes a point cloud filtering algorithm based on image processing. First, the point cloud was preprocessed and mapped onto the image. Second, the noise area was removed from the image and restored to a three-dimensional point cloud by applying image processing. Thus, the point cloud without noise was obtained. This paper analyzes the implementation of the proposed algorithm in detail and verifies the effectiveness and real-time performance of the proposed algorithm based on experimental results.