基于ROI块匹配的全景图像拼接鲁棒性方法

Robust Method for Panorama Mosaic Based on Block Matching in ROI

提出了一种鲁棒性全景图像拼接方法。首先,为了缓和光照条件变化进行直方图均衡化;其次,利用相位相关的方法提取种子点,即重叠区域的高度纹理点,因为感兴趣区域(ROI)被限定在重叠区域,所以能得到更精确的对应点;最后,用加权块匹配算法(BMA)来最小化相机转动产生的图像扭曲问题。通过多种类型全景图像的峰值信噪比(PSNR)与其他算法相比,证明该方法具有很强的鲁棒性。

This paper presents a robust method for constructing a panorama from images. Firstly, in order to mitigate the variation of the illumination conditions, the histograms of the overlapped areas are equalized. Secondly, extracting the seed points, the highly textured point in the overlapped area, by using phase correlation. Because ROI is restricted within overlapped areas, more accurate correspondences are obtained. Finally, the weighted Block Matching Algorithm （BMA） is used to minimize image distortion caused by camera rotation. Comparing the performance of the methods which is Peak Signal to Noise Ratio （PSNR） with multi-types images, results demonstrates the robustness of the method.