基于全局相似性约束的多视点图像拼接

针对在大视差场景下进行图像拼接时容易出现的扭曲、失真、重影等现象，提出了一种基于全局相似性约束的多视点图像拼接算法。在传统细分网格模型的基础上，添加一个先验的全局相似约束项，对待拼接图像在整体上进行相似性转化，同时分别选取视角不同、亮度不同、复杂场景的三组图像，与已有算法的拼接效果进行对比，并采用质量评价方法对图像进行客观评价。结果表明：提出的方法在拼接效果上可以有效降低已有方法对大视差图像拼接时产生的扭曲和重影，且质量评价提高了10%以上；由于对图像的视角及场景深度无严格要求，因而具有良好的适应性，能够更自然地拼接大视差图像。

Multi-viewpoint image stitching based on global similarity constraint

According to the phenomena of image distortion and ghosting that easily occur when an image is stitched in the wide-baseline scene, a stitching method of the multi-viewpoint image based on global similarity constraint is proposed in this paper. Initially, based on the traditional subdivision mesh model, this method added a priori global similarity constraint to make the image perform a similarity transformation as a whole, thus making the image splicing effect more natural. Next, three groups of images with different perspectives, brightness and complex scenes were selected as examples to compare it with the methods of forerunner in their splicing effects and the quality evaluation method was used to objectively analyze the splicing results. The results showed that the method proposed in this paper could reduce the phenomenon of distortion and ghosting of wide-baseline image in splicing effect and the quality evaluation score was improved by more than 10%. No strict requirements were needed for scene depth and visual angle of image. Thus, it had good adaptability and could achieve image stitching naturally in wide-baseline scene.