A Research About Registration Method Based On Multi-feature Fusion

Image registration is always the research emphasis and difficulty in image processing and computer vision,which has wide application in real life.Generally speaking,it is common for industrial image,remote sensing image and medical image.In medical image processing,registration and fusion of different image can integrate their respective information.This will provide more comprehensive information for the doctor and will be of great significance for clinical diagnosis and surgical treatment.In addition,many other techniques depend on the realization of image registration in different degree,such as target detection,motion analysis,image search,image splicing,and so on.In image registration,it includes images with same scenes or different scenes,that is correspondence across scenes.Especially,image registration between images with overlapping region or morphologic variation become more challenging.On account of these circumstances,a registration method based on spatial pyramid grid model and individual entropy correlation coefficient(IECC)is proposed in this paper.First of all,the entire image is divided into specific grid cells in different layers of the pyramid,ranging from an entire image,to coarse grid cells,to each single pixel.According to the locations of SIFT descriptors in different layers,the pyramid grid model synchronously regularizes match consistency at multiple spatial extents.After that,the optimal transformation can be found.Secondly,with the introduction of IECC,the alignment between images will be described more precisely.Then,the more precise registration can be implemented by using b-spline transformation.Experiments show that this proposed method is robust and outperforms state-of-the-art methods in terms of precision.In medical diagnosis and treatment,many problems are involved in carotid artery.To realize high accuracy registration of carotid artery,and then the displacement field can be calculated,an iterative registration method based on shape feature and textural feature is proposed in this paper.It is of great significance in medical diagnosis and treatment.Image matching can be finished by extracting the shape information and using shape context descriptor.Meanwhile,SIFT is robust in target matching,so dense pixel-to-pixel correspondences between two images can be built by getting per-pixel SIFT descriptor.Combining the above,this method can obtain a good registration result.Furthermore,the result will be better after iterative operation.The experiment result proves that the proposed method is robust and has an advantage on accuracy.