Hybrid registration for two-dimensional gel protein images

Image registration technique is fundamental and essential to the accurate and efficient analysis of protein sequence data. But due to the elastic deformations of two-dimensional gel protein eletrophoresis images, their registration still remains a challenge. In this paper, a hybrid 2D gel protein image registration method is proposed. In the first stage of registration, the wavelet-based hierarchical registration which fully exploits the image intensity is used to correct global displacements and affine transformation parameters are achieved. In the second stage, the landmark-based elastic registration is introduced to correct local displacements and enhance registration performance and accuracy. In the proposed method, the hierarchical registration from low resolution to high resolution can accelerate the registration convergence and hence good computational efficiency can be achieved. Though wavelets have been widely used in image processing, its application in the area of gel protein image registration is not well investigated. In this paper, the wavelet-based hierarchical approach is introduced in 2D gel protein image registration. The algorithm makes use of the merits of the existing categories of registration techniques and achieves high performance registration results automatically.