Abstract
Quantitative autoradiography has enormous application potential for many areas of medical research. Three dimensional (3D) reconstruction (representation) of autoradiographic images provides a mechanism for observing and analysing 3D data. Conventional methods, based on tedious manual methods, can sample and analyse only a small portion of this information. Feature correspondence methods in medical image alignment do not perform well for noisy images. Correlation schemes measure the similarities between images; however, they cannot take into consideration size differences. In this paper, we introduce a new approach to image alignment by using disparity function. The function is defined by considering the geometry, and the function parameters are estimated simultaneously from the computation of shape disparity by relaxation method. The decomposition of the disparity function parameters reveals the object rotation, deformation, scaling, and translation in a three dimensional space. The alignment computation is initially based on measured boundary information and intensity changes. Since the computation utilizes all available information and the disparity function is estimated by a least square error criterion, the algorithm is relatively insensitive to noise. 3D reconstructions of two sequences of rat brains are studied, and experiments on alignment and disparities caused by rotation, translation and shape change (deformation and scaling) are presented. The computer reconstruction system provides researchers with a method for 3D data analysis. The scheme can also be employed for other types of medical images.