论文部分内容阅读
As existing methods for infrared image registration and mosaic have some disadvantages at registration accuracy,applicability or computational cost,an effective algorithm based on SURF (Speeded Up Robust Features) features for infrared video mosaic is presented.The approach firstly utilizes SURF algorithm with strong robustness and superior computational performance to extract features and complete feature matching task.Secondly,a RANSAC technique is applied to suppress matching errors,and to estimate the homography between adjacent frames.Finally,an improved global registration strategy is adopted to compute the transformations between video frames and the panorama by a least-square solution,and a panorama from infrared video is obtained using the transformation parameters.The global registration method exploited in our method is distinctly different from conventional technique,which obtains the transformation between video frame and the panorama by means of multiplying transformation matrixes of adjacent frames.A mosaic test was carried out employing infrared video sequence captured by a HJRG-O01 infrared thermal imager,and experimental results indicate that the proposed method is fast and has strong robustness even for infrared images with high noise level,low contrast and few features.Consequently,our method has highly potential capacities for practical application.