摘要
针对传统的图像畸变校正算法建模复杂、实时性差且图像信息易丢失等缺点,提出了一种基于四边形分片逼近控制点的图像畸变校正算法。该方法以标准点阵图像作为量测目标,将数学形态学和滑动邻域操作相结合以确定畸变图像像素点质心,采用基于四边形分片逼近的方法来拟合高次多项式校正模型,运用两步一维线性灰度级插值向后映射算法确定输出图像中像素点的灰度。将该算法在TMS320DM6437DSP上实现,实验结果表明,校正一幅像素为768×494的图像所用的时间为0.036s,畸变校正的误差在0.31个像素以内,有效地避免了边缘信息丢失、空洞及灰度失真现象。
To overcome the shortcomings of traditional distortion correction algorithms, an image distortion correction algorithm based on quadrilateral fractal approach controlling points is proposed. The standard raster image is used as the measurement target in the algorithm, the mathematical morphology is combined with sliding impending domain operation to fix on the distorted image＇s pixel centroid, and the algorithm based on quadrilateral fractal approach controlling points is applied to fit high-order polynomial correction model. For image gray recovery, a two-step one-dimensional linear backward mapping method is used. The algorithm is applied on TMS320DM6437 DSP, and the experimental results show that for a 768 pixels x 494 pixels image, the correction time is 0.036 s. The correction error is within 0.31 pixel and the edge information loss and inanition are effectively avoided.
出处
《光电工程》
CAS
CSCD
北大核心
2009年第5期77-82,共6页
Opto-Electronic Engineering
基金
省部级重大基金资助项目(1020020220606)
关键词
机器视觉
图像畸变
校正
四边形分片逼近
robot vision
image distortion
correction
quadrilateral fractal approach
