Constrained iterative reconstruction by the conjugate gradient method

The conjugate gradient method incorporating the object-extent constraint is applied to image reconstruction of a three-dimensional object using an incomplete projection-data set. The missing information is recovered by constraining the solution with the knowledge of the outer boundary of the object-extent which may be a priori measured or known. The algorithm is derived from the least-squares criterion as an advanced version of conventional iterative reconstruction algorithms such as SIRT (Simultaneous Iterative Reconstruction Technique) and ILST (Iterative Least Squares Technique). In the case of reconstruction from noisy projection data, a method based on the minimum mean-square error criterion is also proposed. Computer simulated reconstruction images of a phantom using limited angle and number of views are presented. The result shows that the conjugate gradient method incorporating the object-extent constraining provides the fastest convergence and the least error.     

Tomographic reconstruction from experimentally obtained cone-beam projections

Direct reconstruction in three dimensions for two-dimensional projection data has been achieved by cone-beam reconstruction techniques. In this paper explicit formulas for a cone-beam convolution and back-projection reconstruction algorithm are given in a form which can be easily coded for a computer. The algorithm is justified by analyzing tomographic reconstructions of a uniformly attenuating sphere from simulated noisy projection data. A particular feature of this algorithm is the use of a one-dimensional rather than two-dimensional convolution function, greatly speeding up the reconstruction. The technique is applicable however large the cone angle of data capture and correctly reduces to the pure fan-beam reconstruction technique in the central section of the cone. The method has been applied to data captured on a cone-beam CT scanner designed for bone mineral densitometry.     

Modified quadratic discriminant functions and the application to chinese character recognition

Issues in the quadratic discriminant functions (QDF) are discussed and two types of modified quadratic disriminant functions (MQDF1, MQDF2) which are less sensitive to the estimation error of the covariance matrices are proposed. The MQDF1 is a function which employs a kind of a (pseudo) Bayesian estimate of the covariance matrix instead of the maximum likelihood estimate ordinarily used in the QDF. The MQDF2 is a variation of the MQDF1 to save the required computation time and storage. Two discriminant functions were applied to Chinese character recognition to evaluate their effectiveness, and remarkable improvement was observed in their performance.