基于CLAHE和改进ZNCC的图像拼接研究

Image Stitching Based on CLAHE and Improved ZNCC

对弱对比度图像进行拼接时,由于对比度较差等原因,待拼接图像上分布的匹配特征点较少,图像配准误差较大。为了解决这一问题,提高图像拼接质量,提出一种基于contrast limited adaptive histogram equalization(CLAHE)和改进zero-mean normalized cross-correlation(ZNCC)的图像拼接算法。在提取特征点前,利用CLAHE算法对弱对比度图像进行预处理,增加图像对比度,增加匹配点数量;然后,使用结合特征点梯度主方向的改进ZNCC算法筛选特征点,提高特征点的正确匹配率;最后,使用筛选后的特征点集合计算变换矩阵,并完成图像拼接。实验结果表明,与其他算法相比,所提算法在弱对比度图像上增加了约25%的正确匹配点,误匹配率相对于SIFT算法降低0.5个百分点~3个百分点,有效提高了图像配准精度,减少了配准重影的出现,优化了图像拼接结果。

When stitching images with weak contrast, there will be a few matching feature points distributed on the images to be stitched because of poor contrast and other factors and the image registration error will be high. To address this problem and improve the quality of image stitching, this study proposes an image stitching algorithm based on contrast limited adaptive histogram equalization(CLAHE) and improved zero-mean normalized cross-correlation(ZNCC). Before feature point extraction, we use the CLAHE algorithm to preprocess the weak contrast image for enhancing the image contrast, which increases number of matching points. Thereafter, the improved ZNCC algorithm combined with the main direction of the gradient of feature points is used to filter feature points, which improves the correct matching rate of feature points. Finally, we use the filtered feature points to calculate the transformation matrix and complete the image stitching. The experimental results indicate that compared with other algorithms, the proposed algorithm increases number of correct matching points by approximately 25% in the weak contrast image and reduces the false matching rate by 0. 5 percentage points-3 percentage points compared with the SIFT algorithm, effectively improving the image registration accuracy, reducing the registration ghosting, and optimizing the image mosaic results.