In order to remove the noise in Low Dose Computed Tomography （LDCT） images and enhance the display effect of the denoised images， an LDCT image enhancement algorithm based on Generative Adversarial Network （GAN） was proposed. Firstly， GAN was combined with perceptual loss and structure loss to denoise the LDCT image. Then， dynamic gray?scale enhancement and edge contour enhancement were performed to the denoised image respectively. Finally， Non?Subsampled Contourlet Transform （NSCT） was used to decompose the enhanced image into multi?directional coefficient sub?images in the frequency domain， and the paired high? and low?frequency sub?images were adaptively fused with Convolutional Neural Network （CNN） to reconstruct the enhanced Computed Tomography （CT） image. Using the real clinical data of the AAPM competition as the experimental dataset， the image denoising， enhancement， and fusion experiments were carried out. The results of the proposed method are 33.015 5 dB， 0.918 5， and 5.99 on Peak Signal?to?Noise Ratio （PSNR）， Structural Similarity Index Measure （SSIM）， and Root Mean Square Error （RMSE） respectively. Experimental results show that the proposed algorithm retains the detailed information of the CT image while removing noise， and improves the brightness and contrast of the image， which helps doctors analyze the patient’s condition more accurately.

In order to solve the problems of too many attempts, slow splicing speed, low restoration accuracy and completeness in artificially restored paper-based cultural relics, an automatic splicing algorithm based on angle and edge length of fragments was proposed. Firstly, the fragment images were pre-processed and coarsely matched according to the angle value of the fragments, and the fragment images with the same angle value were found. Then, on the basis of coarse matching, thin matching was made by using the edge lengths of the angles of the fragments to reduce overlap, and the basic matching results of the fragment images were obtained. Finally, a concave-convex function was used to make up the fragment images of opposite direction, and a oscillating function was used to make up the gap of the final matching images to obtain complete splicing results. Theoretical analysis and splicing simulation experimental results show that compared with automatic splicing algorithms such as feature points, approximate polygon fitting and angle sequence matching, the splicing accuracy, splicing completion and splicing time of the proposed algorithm were improved by at least 12, 11 and 10 percentage points, respectively. The proposed algorithm based on angle and edge features reduces the cumbersome image calculation and accurately corrects the fragment matching result, which enables efficient and highly accurate matching of irregular fragments in actual relic restoration.

To solve the challenge of accurate user group discovering, a user topic discovering algorithm based on trust chain, which was composed by three steps, i.e., topic space discovering, group core user discovering and topic group discovering, was proposed. Firstly, the related definitions of the proposed algorithm were given formally. Secondly, the topic space was discovered through the topic-correlation calculation method and a user interest calculation method for topic space was addressed. Further, the trust chain model, which was composed by atomic, serial, and parallel trust chains, and its trust computation method of topic space were presented. Finally, the detail algorithms of topic group discovering, including topic space discovering algorithm, core user discovering algorithm and topic group discovering algorithm, were proposed. The experimental results show that the average accuracy of the proposed algorithm is 4.1% and 11.3% higher than that of the traditional interest-based and edge density-based group discovering methods. The presented algorithm can improve the accuracy of user group organizing effectively, and it will have good application value for user identifying and classifying in social network.

When the length of the cyclic quorum is the shortest in distributed system, the space complexity and the time complexity of the existing quorum generation algorithm are too high, a new shortest cyclic quorum generation algorithm was presented. Based on the theory of relaxed cyclic difference set, adding elements to the quorum in turn was judged by the condition of maximum allowable number of repetitions. The simulation results show that, when the system node number is 70 to 90, the length of the cyclic quorum is the shortest and the space complexity is O(2n), the time complexity is 3.6E-03 to 6.8E-07 of that of the exhaustive search, and the time complexity of the shortest cyclic quorum generation algorithm is reduced.

A task scheduling strategy based on the dynamically adaptive ant colony algorithm was proposed for the first time to solve the drawbacks like slow convergence and easily falling into local optimal that have long existed in the ant colony algorithm. Chaos disruption was introduced when selecting the resource node, the pheromone evaporation factors were adjusted adaptively based on nodes pheromone and the pheromone were updated dynamically according to the solutions performance. When the number of tasks was greater than 150, compared with the dynamically adaptive ant colony algorithm and ant colony algorithm, time efficiency could be maximally improved up to 319% and resource load was 0.51.The simulation results prove that the proposed algorithm is suitable for improving convergence rate and the global searching ability.

Image layer visual representation has been currently used in computer vision field, but it is difficult for feed-forward image multilayer visual representation methods to deal with local ambiguities. An image multilayer visual representation method based on Latent Dirichlet Allocation (LDA) named LDA-IMVR was proposed. It derived a recursive generative model of LDA by implementing recursive probabilistic decomposition process. Meanwhile, it learned and deduced all layers of the hierarchy together, and improved classification and learning performance by using feed-back style. The approach was tested on Caltech 101 dataset. The experimental results show that the proposed method improves classification performance of objects compared with related hierarchical approaches, and it achieves better results in learned components and image patches visualization.

It is necessary to generate the quorums as soon as possible in large-scale fully distributed system for its mutual exclusion problem. Based on the theory of relaxed cyclic difference set, the definition of second relaxed cyclic difference set was proposed. After researching the new concepts, the subtraction steps in previously classical methods can be changed into summation steps. Furthermore, a lot of summation steps can be cut down by the recurrence relation deduced from the summation steps. The time complexity of this algorithm is just only O(n) and the size of the symmetric quorums is still close to 2n^(1/2).

How to reduce the time complexity of the quorum generation algorithm effectively when the quorum length does not increase significantly is a question must be resolved to all researchers of symmetric quorum generation algorithm for distributed mutual exclusion. A new quorum generation algorithm was proposed in this paper by adopting the local recursion. This algorithm can reduce the time complexity of the quorum generation algorithm effectively and ensure the quorum length not increasing significantly than WK's algorithm and global recursion algorithm. Therefore, through researching the features of quorum, the contradiction between quorum length and time complexity of the quorum generation algorithm can be improved.

Multi-instance neural networks are a kind of neural networks to solve multi-instance learning problems.There is a function that is not differentiable in multi-instance neural networks,so the predictive accuracy is low if multi-instance neural networks are trained with the back-propagation approach.In order to achieve higher predictive accuracy,an improved genetic algorithm for optimizing multi-instance neural networks was presented.Convergence rate was increased and premature convergence was overcome by means of local search operator,suppress operator and adaptive calculations of probabilities for operators.Some experiments on well-known test data show that multi-instance neural networks that are optimized by the improved genetic algorithm heighten significantly predictive accuracy and computational expensiveness of the algorithm is less than other algorithms.