The Study Of Image Registration Method Based On NSCT

With the rapid development of information technology and the increasing levels of technology, image registration technique in the field of image processing has become an extremely important technology, and has become the hot topic of the research for the researchers. Image registration is a process to establish the corresponding relationship between two or more images, to determine the corresponding geometric transformation parameters, then, in order to convert the two or more images to the same coordinate system, to achieve the best matching in the pixel level.Image registration methods can be divided into image registration methods based on gray information and image registration methods based on feature. Due to the multi-resolution and localization of wavelet transform, feature points detection based on wavelet transform has been widely applied in the field of image processing, such as image registration, pattern recognition, computer vision and so on. Especially, method based on wavelet transform modular maximum is commonly used to match the images. But aiming at solving the problem of the poor matching results for multi-modal image, the optimized wavelet transform modular maximum method is proposed to extract feature points, and it is applied to the infrared and visible image matching, CT and MR image matching.However, due to the limited ability of wavelets in two dimensions to capture directional information, the multi-scale and multi-direction NSCT transformation that can capture the intrinsic geometrical structures has been considered recently. This article focuses on putting forward a new registration method, which combines the multi-direction character of NSCT transformation with the fast speed of DAISY descriptor, and it is applied to the infrared/visible image matching. Visual observation method and accuracy comparison method have been used to evaluate the performance. Several experiments conducted illustrate that this method in this paper has a good performance of efficiency, accuracy, and it can also be applied to many situations for some multi-modal images.