多曝光图像移动边界伪影自主校正方法仿真

Multi-exposure image moving boundary artifact autonomous correction method simulation

针对传统方法存在伪影校正时间较长、标准差较高、图像的分辨率较低等问题,提出一种基于改进Canny的多曝光图像移动边界伪影自主校正方法。根据样本空间中马氏距离不稳定的特点计算特征空间中的马氏距离,对图像相关的数据进行相关性分析以及降维处理,提取数据的主成分。对马氏距离的计算方法进行简化,通过马氏距离获取高斯加权函数对图像进行去噪。在经过去噪的图像上,利用S-L滤波器对原始投影数据进行滤波处理,增强伪影信息。设定角度阈值对梯度方向进行限制,排除斜角方向边缘的检测,实现竖直方向边缘检测。在上述基础上,设定梯度阈值以及链长度阈值完成伪影边缘点的检测与连接,引用分段B样条拟合法对伪影进行校正。实验结果表明,所提方法能够有效减少图像校正时间,并且能够较好的保存图像的细节以及提高分辨率。

This research proposes an artifact autonomous correction method of moving boundary of multi-exposure image based on the modified Canny. The research calculated mahalanobis distance in feature space according to characteristic of unstable mahalanobis distance in sample space, then carried out correlation analysis and dimensionality reduction for relevant data of image to extract principal component of data. The research simplified computing method of the mahalanobis distance and acquired Gaussian weighted function via the mahalanobis distance to eliminate noise of the image. The research used S-L filter to carry filtering process for original projection data in the de-noised image to enhance artifact information. The research limited gradient direction and eliminated detection of edge of oblique angle direction via setting up angle threshold to achieve edge detection of vertical direction. On that basis, our research set up gradient threshold and link length threshold to complete detection and connection of artifact edge. Finally, the research introduced segmentation B-spline fitting method to correct the artifact. Simulation results show that the method can reduce the correction time effectively. Meanwhile, the method can preserve detail of image preferably and improve resolution.