In high-dynamic-range (HDR) imaging, a radiance map of a HDR scene can be constructed by capturing the scene multiple times with a digital camera at different exposure settings and then digitally combining the images. As we show in this paper, the signal-to-noise ratio (SNR) of the resulting HDR radiance map depends strongly on the number of images that are captured. We present an analytical model for computing SNR as a function of the set of exposure settings used in image capture and the physical noise parameters of the digital camera. We find that the average dynamic range of the HDR image varies as the inverse of the square-root of the inter-exposure spacing. We also find that using a denser exposure set can significantly reduce the inter-pixel variability and spatial variability of SNR in the final HDR radiance map.

This paper considers the problem of high-dynamic-range (HDR) image capture using low-dynamic-range (LDR) cameras. We present three different minimal-bracketing algorithms for computing minimum-sized exposure sets bracketing of HDR scenes. Each algorithm is applicable to a different HDR-imaging scenario depending on the amount of target-scene-irradiance information and real-time image processing available at the time of image acquisition. We prove the optimality of each algorithm with respect to its ability to obtain a theoretically minimum-size bracketing set of exposures. We also provide closed-form expressions for computing minimal-bracketing exposure sets for two common types of HDR-imaging systems, those with geometrically varying and arithmetically varying exposure settings. We experimentally demonstrate the advantages of the proposed methods by capturing and processing multiple HDR scenes using minimal-bracketing and 1-stop bracketing methods. The results show that minimal-bracketing can be used to produce high-quality HDR images, while requiring only one third as many LDR images be acquired compared to 1-stop bracketing. We also perform a detailed SNR analysis that quantifies the tradeoff between signal-to-noise ratio and image-bracketing-set size.

In this study, we examine control-plane throughput in optical burst switching (OBS) networks with electronic header processing. In a large-scale OBS network, headers arrive to the burst scheduler at a very high rate. This can potentially cause overloading of the core-node control-processor and lead to data loss if headers are not processed in a timely manner. In this work, we examine the effect of control-plane congestion on the overall loss performance of OBS networks. We present preliminary results that show how this congestion can become the primary source of loss and can ultimately limit the throughput and the scalability of the OBS system if not properly accounted for when designing the OBS network.

In this paper, we introduce a new signalling architecture called Dual-header Optical Burst Switching (DOBS) for next generation burst-switching optical networks. DOBS decouples the resource reservation process from the service request process in core nodes and allows for delayed scheduling to be implemented. This relaxes the constraints on burst scheduling operations and allows the offset sizes of bursts to be precisely controlled in core nodes without the use of fiber delay line buffers. This allows for increased flexibility, control, and performance. To demonstrate the benefit of delayed scheduling and core. node offset control, we examine the performance of a DOBS system in which the offset size of every burst on a core link is set to a constant value. Using simulation and analysis, we show that the resulting constant-scheduling-offset (CSO) system realizes lower ingress delay, higher throughput, and better fairness than conventional single-header OBS systems, while simultaneously requiring only O(1) burst scheduling complexity. In a 16-channel system with full wavelength conversion and no fiber delay line buffers, the CSO DOBS system achieved a blocking probability 50% lower than that of a similar LAUC-VF JET OBS system. The CSO DOBS system also achieved perfect fairness, both with respect to burst length and with respect to the residual path length of bursts

In this paper we introduce a new signalling architecture called dual-header optical burst switching (DOBS) for next generation burst-switching optical networks. Using DOBS, the functional offset size of every burst on a given link can be set to the same size without the use of fiber delay line buffers. This allows DOBS to realize lower burst-scheduling complexity, lower ingress delay, higher throughput and better fairness than conventional single-header OBS systems. We present a new burst-scheduling algorithm called free channel queue scheduling that requires only O(1) time to execute and that achieves optimal performance in constant-offset DOBS systems. Using simulation, we find that the blocking probability of a 16-channel DOBS system is 50% lower than that of a similar LAUC-VF JET OBS system. We also show that DOBS achieves better fairness than JET OBS with respect to burst length and with respect to the residual path length of bursts.

In this paper, we present for the first time an analytical model that quantifies the mechanism by which offset size affects priority in multiclass optical-burst switching (OBS) systems. Using the model, we derive an exact expression for the distribution of the number of bursts that contend with an arriving burst. The model is applicable to systems in which each class has an arbitrary burst-length distribution and an arbitrary offset size. We also derive accurate approximate expressions for the burst-blocking probability of premium-class traffic, as well as expressions for the sensitivity of premium-class performance to offset jitter and variations in the arrival rates of each class. In a case study, we find that scaling up a system in terms of the number of wavelengths and the traffic load significantly improves not only the burst-blocking performance of the premium class, but also its sensitivity to lower class traffic variations. We also use the model to dimension and provision the system to guarantee a minimum level of premium-class blocking and premium-class robustness to low-class load variations.

In optical burst switching (OBS) networks, class differentiation and isolation can be achieved by assigning adequately long time offsets between the control packet and payload of high-class bursts. While it has been recognized that the length distribution of low-class bursts plays a role in determining the size of the offsets required, there have been no studies on the effect of other factors that may also be significant. In this paper we examine the effect of the ratio of the arrival rates of low-class and high-class traffic on the level of isolation achieved in OBS networks with quality of service offsets. We show that the level of isolation in the network depends on the arrival rate of low-class traffic, especially when the amount of low-class and high-class traffic in the system is comparable. When we vary the ratio between low and high-class arrival rates from 0.1 to 10, an additional offset of three times the mean low-class burst length is required to achieve the same level of isolation. These results imply that it is important for researchers and network designers to take into account the amount of low-class traffic in the network when provisioning offsets for class differentiation in OBS networks.

This paper presents an analytical model that evaluates the blocking probability of each service class in optical burst-switching networks. The model is applicable to systems with arbitrary burst length distributions and arbitrary-sized QoS header offsets. Thus, unlike previous models, it is applicable to the design and study of networks with a wide range of traffic characteristics, including systems in which higher classes are not necessarily isolated from lower classes and systems in which the conservation law does not necessarily hold. We derive explicit expressions for blocking probability both the cases of constant burst lengths and exponentially distributed burst lengths and verify the model's accuracy through simulation. We show the model to be accurate for a number of different traffic loads and class priorities. For an OBS system with two classes and a 1:10 ratio of high-priority to low-priority traffic, our model is able to predict accurately the blocking probability for each class, whereas the predictions from a model that assumes isolation deviates by as much as an order of magnitude from the simulation results for the higher priority class.

In this paper, we examine the use of limited tunability in reconfigurable optical add-drop multiplexers (L-ROADM). L-ROADMs can add or drop from only a subset of adjacent wavelengths on the network and are less costly than fully-reconfigurable optical add-drop multiplexers (F-ROADMs). We quantify the trade-off between tuning range and the number of F-ROADMs that can be replaced by L-ROADMs without sacrificing the set of connections that can be established. For the limited-add and drop case, an analytical solution for the band size is found. For the limited-add or limited-drop case, a nearly linear relationship was found between the size of the band and the number of L-ROADMs required, and the number of additional wavelengths required never exceeded 20%. For example, if half of the nodes in the ring were equipped with L-ROADMs that operated on 50% of the total spectrum, full connectivity could still be achieved by employing as few as 7% extra wavelengths.

In all-optical wavelength-switched WDM networks, 3R regeneration does not need to be performed at every node in a path. Thus, a significant cost savings can be realized by efficiently provisioning a limited amount of 3R regeneration resources in each node in the network and utilizing these resources efficiently. In order to aid in these two tasks, we present an analytical model to predict the location and relative frequency of 3R regeneration requests in the network. Because the model is based solely on the topological information of the network, the predictions provided are very general and are independent of specific network operating parameters, such as the routing protocol employed. Simulation results are also presented to verify the accuracy of the model.