Studies Of Technologies And Methods On Image Mosaic

The image or video with wide view and high resolution becomes more and more important in some domains,for example,the panoramic view in the photogrammetry and the panoramic video supervisory system,especially the hypothesized environment in the virtual reality technology and so on. Because the image mosaics technology has solved the contradictory question between the field of vision and the resolution,it becomes a hotspot question in the image processing domain.This thesis discusses the stitching of multiple images taken by ordinary camera or other hand-by devices like mobile phone. For the character that the camera shake big and the image rotate big, This thesis improve image registration algorithm ,effectively improve the gray image registration speed and Realized the color image registration .Base on the shortage of the SURF, an improved image registration algorithm based SURF point is presented. This algorithm first extracts the SIFT feature points, and then detects the Harris points in the neighbour district of each SURF feature point. So the unique character of the feature points will be boosted up. Because SURF feature is invariant to the rotation and scale, as well as it is robust to added noise, viewpoint change and illumination change. It is possible to achieve image registration successfully when large scale change occurs. Meantime, the SURF feature point extraction method is improved by reducing some unstable feature points in advance, so that the speed of matching and the precision of correct matching are improved. Experimental results show that the proposed algorithm is invariant to rotation, scale change, and it has good robustness to noise and intensity change. Besides, the speed of matching is improved by about 2 times.A novel Harris-based registration method for color images is proposed. Extending differential-based operations to color images is hindered by the multi-channel nature of color images. The derivatives in different channels can point in opposite directions, hence cancellation might occur by simple addition. We combine feature detectors with photometric invariance theory to construct invariant features. Experiments show that color features perform better than luminance based features and that the additional photometric information is useful to discriminate between different physical causes of features.