In this paper, we present a non-linear (multi-affine) registration algorithm based on a local polynomial expansion model. We generalize previous work using a quadratic polynomial expansion model. Local affine models are estimated using this generalized model analytically and iteratively, and combined to a deformable registration algorithm. Experiments show that the affine parameter calculations derived from this quadratic model are more accurate than using a linear model. Experiments further indicate that the multi-affine deformable registration method can handle complex non-linear deformation fields necessary for deformable registration, and a faster convergent rate is verified from our comparison experiment.

This paper presents a novel method for assisting surgeons in automatically computing an optimal surgical plan by directly specifying the desired correction to a facial outline. First, the desired facial appearance is designed using a 3D sculpturing tool, while the cut regions of the skull are defined based on facial anatomy. Then, the deformation of the face meshes is performed using an improved biomechanical model in which virtual external forces are driven by the displacements corresponding to the differences of node coordinates between the original and specified face meshes, and free nodes and fixed nodes are defined in terms of the contact surfaces between the soft tissues and the bones within the cut regions. Finally, the shape of the contact surfaces is updated following the deformation of the soft tissues. After registering the deformable contact surfaces and the cut surfaces, the final positions of the cut bones are estimated. Evaluation of preliminary experimental results quantitatively shows the effectiveness of the proposed approach.

Finger vein recognition is a biometric technique which identifies individuals using their unique finger vein patterns. It is reported to have a high accuracy and rapid processing speed. In addition, it is impossible to steal a vein pattern located inside the finger. We propose a new identification method of finger vascular patterns using a weighted local binary pattern (LBP) and support vector machine (SVM). This research is novel in the following three ways. First, holistic codes are extracted through the LBP method without using a vein detection procedure. This reduces the processing time and the complexities in detecting finger vein patterns. Second, we classify the local areas from which the LBP codes are extracted into three categories based on the SVM classifier: local areas that include a large amount (LA), a medium amount (MA), and a small amount (SA) of vein patterns. Third, different weights are assigned to the extracted LBP code according to the local area type (LA, MA, and SA) from which the LBP codes were extracted. The optimal weights are determined empirically in terms of the accuracy of the finger vein recognition. Experimental results show that our equal error rate (EER) is significantly lower compared to that without the proposed method or using a conventional method.

This paper tackles the problem of video concept detection using the multi-modality fusion method. Motivated by multi-view learning algorithms, multi-modality features of videos can be represented by multiple graphs. And the graph-based semi-supervised learning methods can be extended to multiple graphs to predict the semantic labels for unlabeled video data. However, traditional graphs represent only homogeneous pairwise linking relations, and therefore the high-order correlations inherent in videos, such as high-order visual similarities, are ignored. In this paper we represent heterogeneous features by multiple hypergraphs and then the high-order correlated samples can be associated with hyperedges. Furthermore, the multi-hypergraph ranking (MHR) algorithm is proposed by defining Markov random walk on each hypergraph and then forming the mixture Markov chains so as to perform transductive learning in multiple hypergraphs. In experiments on the TRECVID dataset, a triple-hypergraph consisting of visual, textual features and multiple labeled tags is constructed to predict concept labels for unlabeled video shots by the MHR framework. Experimental results show that our approach is effective.

Multi-objective differential evolution (MODE) is a powerful and efficient population-based stochastic search technique for solving multi-objective optimization problems in many scientific and engineering fields. However, premature convergence is the major drawback of MODE, especially when there are numerous local Pareto optimal solutions. To overcome this problem, we propose a MODE with a diversity enhancement (MODE-DE) mechanism to prevent the algorithm becoming trapped in a locally optimal Pareto front. The proposed algorithm combines the current population with a number of randomly generated parameter vectors to increase the diversity of the differential vectors and thereby the diversity of the newly generated offspring. The performance of the MODE-DE algorithm was evaluated on a set of 19 benchmark problem codes available from http://www3.ntu.edu.sg/home/epnsugan/. With the proposed method, the performances were either better than or equal to those of the MODE without the diversity enhancement.

In this paper, retransmission strategies of the network-coding-based packet network are investigated. We propose two retransmission strategies, the packet-loss-edge-based retransmission strategy (PLERT) and the minimum retransmission strategy (MRT), which focus on optimizing the retransmission efficiency without the constraint on the encoding field size. We compared the performances of the proposed retransmission strategies with the traditional automatic repeat-request (ARQ) strategy and the random retransmission strategy. Simulation results showed that the PLERT strategy works well when the packet loss rate is small. Among these retransmission strategies, the performance of the MRT strategy is the best at the cost of the high complexity that is still polynomial. Furthermore, neither of the proposed strategies is sensitive to the encoding field size.

A generalized single-phase multilevel current source inverter (MCSI) topology with self-balancing current is proposed, which uses the duality transformation from the generalized multilevel voltage source inverter (MVSI) topology. The existing single-phase 8- and 6-switch 5-level current source inverters (CSIs) can be derived from this generalized MCSI topology. In the proposed topology, each intermediate DC-link current level can be balanced automatically without adding any external circuits; thus, a true multilevel structure is provided. Moreover, owing to the dual relationship, many research results relating to the operation, modulation, and control strategies of MVSIs can be applied directly to the MCSIs. Some simulation results are presented to verify the proposed MCSI topology.

Scheduling projects at the activity level increases the complexity of decision making of project portfolio selection but also expands the search space to include better project portfolios. An integer programming model is formulated for the project portfolio selection and scheduling problem. An iterative multi-unit combinatorial auction algorithm is proposed to select and schedule project portfolios through a distributed bidding mechanism. Two price update schemes are designed to adopt either a standard or an adaptive Walrasian tatonnement process. Computational tests show that the proposed auction algorithm with the adaptive price update scheme selects and schedules project portfolios effectively and maximizes the total net present value. The price profile generated by the algorithm also provides managerial insights for project managers and helps to manage the scarce resources efficiently.