Images acquired from digital cameras are usually interfered by smoke, which may degrade the performance of object detection. There are few algorithms focused on smoke removal for still images so far and we usually use haze removal algorithms to remove smoke instead. However, there exist some differences between haze and smoke (e.g. particle properties and localization). Thus, a dehaze algorithm usually has limited performance for smoke removal. In this paper, we propose a novel smoke removal algorithm based on machine learning and smoke detection techniques. Moreover, we observed that the intensity distributions are not the same for different color channels in smoky images. Therefore, the proposed algorithm trains the models corresponding to each color channel and remove smoke from RGB channels separately. Simulations show that the proposed algorithm can significantly remove smoke. Moreover, as far as we know, the proposed algorithm is the first smoke removal algorithm for static images.

We propose the application of cosine windowing for the orthogonal frequency-division multiplexing (OFDM) systems to self-cancel intercarrier interference (ICI) in fast time-varying channels prior to receptions. With a time-domain cosine window immediately after the inverse discrete Fourier transform (IDFT) unit in OFDM transmitters, the ICI fractions from adjacent subcarriers significantly cancel one another at the expense of the orthogonality violation among subcarriers in the main lobe. As a result, the frequency-domain channel matrix reshaped by the cosine windowing can be closely approximated to a strictly banded matrix. In the complex exponential basis expansion model (CE-BEM), we present the estimation of the channel matrix with the assistance of the pilot clusters. Simulation results show that the receivers implementing the CE-BEM channel estimation and the low-complexity block turbo minimum mean square error (MMSE) equalization perform with considerably lower bit error rates (BER) even in very fast time-varying channels.

For the obstacle-avoiding rectilinear Steiner minimal tree (OARSMT) problem, we present a Steiner-point-based algorithm that achieves the best practical performance among existing heuristics. We first propose a new concept of Steiner point locations, creating a linear-space routing graph with satisfactory Steiner point candidates to resolve the bottleneck of most existing heuristics. Then, we propose a Steiner-point-based framework to yield a solution, which is close to the key to the handling of the OARSMT problem. Experimental results show that this algorithm achieves excellent solution quality and speed performance at the same time. We also extend the Steiner-point-based framework to the obstacle-avoiding preferred direction Steiner tree problem with a good performance.

The ad hoc sensor network has been widely applied to various applications, such as environmental data collection and surveillance. To enable these applications, the accurate localization of sensor nodes is crucial. The DV-Hop provides a basic scheme to retrieve the localization information without GPS. The DV-Hop scheme requires the anchor nodes to be localized in advance, and the locations of the anchor nodes are used to localize other unknown nodes. The hop count between two anchor nodes can be exchanged through multi-hop routing. Using the hop counts and locations of other anchor nodes, the anchor nodes obtain the average distances per hop among one another. The distance estimation errors are caused by the uncertainties in per-hop distance estimation and the communication ranges. We propose a scheme where a node ranks the orders of its neighbor nodes through exchanging neighbor information locally. The neighbor information with orders provides useful information for node localization. Besides, the analysis reveals the probability among the order distance and the node density. Therefore, the distance estimation is improved by using the order information of neighbor nodes. The simulation results show more than 22% error reduction compared with the DV-Hop.

For the obstacle-avoiding rectilinear Steiner minimal tree problem, this paper presents an O(n log n)-time algorithm with theoretical optimality guarantees on a number of specific cases, which required O(n3) time in previous works. We propose a new framework to directly generate O(n) critical paths as essential solution components, and prove that those paths guarantee the existence of desirable solutions. The path-based framework neither generates invalid initial solutions nor constructs connected routing graphs, and thus provides a new way to deal with the OARSMT problem. Experimental results show that our algorithm achieves the best speed performance, while the average wirelength of the resulting solutions is only 1.1% longer than that of the best existing solutions.

For the obstacle-avoiding rectilinear Steiner minimal tree (OARSMT) problem, this paper presents a Steiner-point based algorithm to achieve the best practical performance in wirelength and run time. Unlike many previous works, the Steiner-based framework is more focused on the usage of Steiner points instead of the handling of obstacles. This paper also proposes a new concept of Steiner point locations to provide an effective as well as efficient way to generate desirable Steiner point candidates. Experimental results show that this algorithm achieves the best solution quality in Θ (n log n) empirical time, which was originally generated by applying the maze routing on an Ω(n2)-space graph. The Steiner-point based framework and the new concept of Steiner point locations can be applied to future research on the OARSMT problem and its generations, such as the multi-layer OARSMT problem.

Rectilinear Steiner trees are used to route signal nets by global and detail routers in VLSI design for a long time. However, in current IC industry, there are significantly increasing obstacles to be considered, such as large-scale power networks, pre-routed nets, IP blocks, and antenna jumpers. Accordingly, the obstacle-avoiding rectilinear Steiner minimal tree (OARSMT) problem has become more important. In this article, we propose a new routing graph, obstacle-avoiding routing graph (OARG), for the OARSMT problem. Due to the important properties of OARG, we construct a 3-step algorithm and a local refinement scheme, which both can take advantage of these properties, to find a suboptimal solution efficiently. Furthermore, each step of our 3-step algorithm as well as the local refinement scheme has theoretical or practical benefits. Therefore, each of them can be applicable to other existing works for general or specific considerations such as efficiency or effectiveness. Extensive experimental results show that our method outperforms all existing works in terms of wirelength and achieves the best speed performance.

In IC design, rectilinear Steiner trees have been used to route signal nets by global and detail routers for a long time. Recently, there are more complicated processing conditions for routing to be considered, such as multiple routing layers, obstacles, and preferred directions. Furthermore, routability is also an important issue for modern routing which handles more than ten thousand signal nets. As a result, how to meet the processing conditions and consider the routability at the same time is becoming important. In this paper, we formulate a routing problem, called the obstacle-avoiding preferred direction Steiner tree (OAPDST) problem, which can deal with more practical processing conditions and achieve acceptable routability. To the best of our knowledge, this is the first attempt to formulate this problem. Then, we propose a routing graph, called preferred direction evading graph (PDEG), for this problem, and prove that at least one optimal solution can be found on PDEG. As a result, by using PDEG as the solution space, more efficient and effective methods can be found for the OAPDST problem. Based on PDEG, we also construct an approximation algorithm for the OAPDST problem to provide stable and effective solutions. Experimental results show that our method can perform well for the OAPDST problem

The electrostatic discharge (ESD) problem has become a challenging reliability issue in nanometer-circuit design. High voltages that resulted from ESD might cause high current densities in a small device and burn it out, so on-chip protection circuits for IC pads are required. To reduce the design cost, the protection circuit should be added only for the IC pads with an ESD current path, which causes the ESD current path analysis problem. In this paper, we first introduce the analysis problem for ESD protection in circuit design. We then model the circuit as a constraint graph, decompose the ESD connected components (ECCs) linked with the pads, and apply breadth-first search (BFS) to identify the ECCs in each constraint graph and, thus, the current paths. Experimental results show that our algorithm can very efficiently and economically detect all ESD paths. For example, our algorithm can detect all ESD paths in a circuit with more than 1.3 million vertices in 1.39 s and consume only 44-MB memory on a 3.0-GHz Intel Pentium 4 PC. To the best of our knowledge, our algorithm is thefirstpointtoolavailable to the public for the ESD analysis.

Peer-to-peer (P2P) communications have attracted a great deal of attention from the network research community in recent years. However, due to the fundamental limitations of wireless environments, providing reliable data availability for P2P applications over wireless ad hoc networks is still a major challenge. To address the problem, a distributed and randomised scheme based on self-avoiding walks is proposed. The scheme concatenates disparate network layers, with the goal of recovering from routing failures that disrupt P2P data accessibility. In addition, a probabilistic approach is presented that explores the tradeoffs between several system parameters. Some new analysis tools, such as path coupling, are utilised which provide a better understanding of the system's operations. That the proposed concepts and techniques make a significant contribution to the design of effective and efficient P2P applications in wireless ad hoc networks is believed.

Rather than conventional segment partition methods in link/path/sub-path based protection switching schemes, we discuss the multi-metrics resource-based network partition approaches to meet the requirement of group switching and adapt to dynamic traffic demands. By the route selection step according to exchanged capacity estimation information, group reconfiguration process and rerouting in each group, channel utilization of network will be maximized and hence the overall quality of protection can be achieved. We consider both the static and dynamic group partition strategies under dynamic traffic demands and examine their relative influence of various network metrics. The simulation results show that resource utilization can be optimized at the expense of moderate restoration penalty under heavy load in the dynamic traffic environment over dense mesh backbone networks.

This paper presents an efficient technique to map the minimum vertex cover and two closely related problems (maximum independent set and maximum clique) onto the Hopfield neural networks. The proposed approach can be used to find near-optimum solutions for these problems in parallel, and particularly the network algorithm always yields minimal vertex covers. A systematic way of deriving energy functions is described. Based on these relationships, other NP-complete problems in graph theory can also be solved by neural networks. Extensive simulations were performed, and the experimental results show that the network algorithm outperforms the well-known greedy algorithm for vertex cover problems.