CAJ | 학술논문

针对未标定的立体图像对进行校正。在该方法中，相机不需要预先进行标定，只需要提取立体图像上的几组匹配点，利用匹配点在校正后也满足极线几何约束这一条件，最小化所有匹配点的桑普森误差值，即可得到系统方程，通过求解系统方程得到校正变换矩阵，进而完成立体图像对的校正。实验表明，经过校正后的立体图像对的尺寸基本保持不变，垂直视差值被消除，因此后续的立体匹配完全可以在校正后图像的水平扫描线上进行。
침대미표정적입체도상대진행교정。재해방법중，상궤불수요예선진행표정，지수요제취입체도상상적궤조필배점，이용필배점재교정후야만족겁선궤하약속저일조건，최소화소유필배점적상보삼오차치，즉가득도계통방정，통과구해계통방정득도교정변환구진，진이완성입체도상대적교정。실험표명，경과교정후적입체도상대적척촌기본보지불변，수직시차치피소제，인차후속적입체필배완전가이재교정후도상적수평소묘선상진행。
In this paper we focus on the rectification of uncalibrated stereo images .In our approach , the cameras don't need to be calibra-ted beforehand , we only extract a few of matched points in stereo image .By making use of the condition that the matched points in rectified images meet the epipolar geometry constraint as well , we minimise the Sampson error value of all the matched points , and then obtain the sys-tem equations .Through solving the system equations , we achieve the rectification of the stereo image .Experiments show that , the size of the rectified stereo images pair keeps almost unchanged , and the vertical disparity is eliminated , so the subsequent stereo matching can be ful-filled on the horizontal scanning line of the rectified images completely .