点是否在矩形中判断 改进投影变换和保留结构特征的拼接图像修复算法

魏启元+吕晓琪+谷宇

摘 要： 拼接图像含有不规则的边界，需通过裁剪图像和补充像素获得矩形图像。直接裁剪的方式丢弃了图像周围的像素；基于样本块的图像修复算法存在物体结构上的不连续和不完整；利用马尔科夫随机场模型修补图像的方法搜索范围大，效率低。提出通过改进投影变换实现拼接图像的空间变换，减少拼接后图像周围的空缺区域面积，结合边界保留的细缝段裁剪算法对拼接图像进行矩形化扩充，实现拼接图像周围像素和内部物体结构信息的保留。点是否在矩形中判断实验结果表明，该方法实现了拼接图像视角最大化，未引入视觉可见形变，有效实现了拼接图像的矩形修复。

关键词： 拼接图像； 矩形化扩充； 细缝段裁剪； 边界保留； 图像修补

中图分类号： TN911.73?34； TP391 文献标识码： A 文章编号： 1004?373X（2017）23?0038?05

Abstract： The splicing image contains the irregular boundary， and its rectangular shape is acquired by means of image cutting and pixel supplementary。 Since the direct cutting mode may discard the pixels around the image， the image completion algorithm based on sample block has the discontinuous and incomplete object structure， and the image completion method based on Markov random field model has broad search scope and low efficiency， a method of improving the projection transformation is proposed to realize the space transformation of the splicing image， so as to reduce the vacant areas around the spliced image， and combined the seam segment carving algorithm for boundary feature reservation to perform the rectangular expansion for the splicing image so as to realize the reservation of the pixels around the splicing image and object structure information inside the image。点是否在矩形中判断

The experimental results show that the proposed method can realize the maximum visual angle of the splicing image， and rectangular completion of the splicing image without the visual visible deformation。

Keywords： splicing image； rectangular expansion； seam segment carving； edge preservation； image completion