基于3DGMM的多视点视频直方图颜色校正

多视点视频在采集过程中,由于受到摄像机所处位置以及摄像机自身的参数设置等因素的影响,会造成视点间颜色的不一致,针对这一情况,提出基于三维高斯混合模型的多视点视频直方图颜色校正算法。算法首先获取源图像以及目标图像的直方图并检测直方图的峰值个数,然后结合三维高斯混合模型对源图像以及目标图像进行聚类分析,获取中心矢量,最后计算两图像中像素点距离中心矢量的最小欧几里得距离并分割图像,对源图像和目标图像的匹配区域进行直方图匹配校正。实验结果表明,该算法与传统直方图匹配算法相比,避免了校正过度的影响,校正效果较好。

Multi-view Video Color Correction Method Based on 3DGMM

In the process of multi-view video capturing, the camera location and parameters setting will lead to the color inconsistent between different viewpoints. In this paper, a new multi-view video histogram color correction algorithm based on 3D Gaussian Mixture Models (3DGMM) is proposed. Firstly, the cumulative histogram of source image and target image are obtained and the peak number of the histogram is calculated by the method. For the clustering analysis of source image and target image, the three-dimensional Gaussian Mixture Models is performed, while the center vector is obtained. Finally, the algorithm calculates the minimum Euclidean distance between each pixel and the center vector to segment the image, histogram matching correction is followed. Experimental results show that the proposed method avoids the excessive correction and gets a good effect comparing with the traditional histogram matching algorithm.