Since real imaging systems are imperfect, the acquired images are always distorted by blur, scale... more Since real imaging systems are imperfect, the acquired images are always distorted by blur, scale and rotation transformation. Then the registration of these degraded images has become an important task in many applications in which the moment invariants are usually efficient tools. However, the existing methods can only deal with the slightly distorted images and have the problems of information redundancy. Besides, some methods have overlapping constraint that the images to be aligned should be fully included into the reference images. In this paper, we proposed a novel method in which a new set of combined invariants based on Legendre moment holding for blur, rotation and scale degradation simultaneously were constructed as feature descriptors, and scale-invariant Harris-Laplace detector was applied to exact feature points. The experimental results show that our method can work well without overlapping constraint, especially when the distortion is great.
Since real imaging systems are imperfect, the acquired images are always distorted by blur, scale... more Since real imaging systems are imperfect, the acquired images are always distorted by blur, scale and rotation transformation. Then the registration of these degraded images has become an important task in many applications in which the moment invariants are usually efficient tools. However, the existing methods can only deal with the slightly distorted images and have the problems of information redundancy. Besides, some methods have overlapping constraint that the images to be aligned should be fully included into the reference images. In this paper, we proposed a novel method in which a new set of combined invariants based on Legendre moment holding for blur, rotation and scale degradation simultaneously were constructed as feature descriptors, and scale-invariant Harris-Laplace detector was applied to exact feature points. The experimental results show that our method can work well without overlapping constraint, especially when the distortion is great.
Uploads
Papers by Xiubin Dai