Arun N. Netravali

Picture coding: A review

This paper presents a review of techniques used for digital encoding of picture material. Statistical models of picture signals and elements of psychophysics relevant to picture coding are covered first, followed by a description of the coding techniques. Detailed examples of three typical systems, which combine some of the coding principles, are given. A bright future for new systems is forecasted based on emerging new concepts, technology of integrated circuits and the need to digitize in a variety of contexts.

Motion and structure from feature correspondences: a review

We present a review of algorithms and their performance for determining three-dimensional (3D) motion and structure of rigid objects when their corresponding features are known at different times or are viewed by different cameras. Three categories of problems are considered, depending upon whether the features are two (2D) or three-dimensional (3D) and the type of correspondence: a) 3D to 3D (i.e., locations of corresponding features in 3D space are known at two different times), b) 2D to 3D (i.e., locations of features in 3D space and their projection on the camera plane are known, and c) 2D to 2D (i.e., projections of features on the camera plane are known at two different times). Features considered include points, straight lines, curved lines, and corners. Emphasis is on problem formulation, efficient algorithms for solution, existence and uniqueness of solutions, and sensitivity of solutions to noise in the observed data. Algorithms described have been used in a variety of applications. Some of these are: a) positioning and navigating 3D objects in a 3D world, b) camera calibration, i.e., determining location and orientation of a camera by observing 3D features whose location is known, c) estimating motion and structure of moving objects relative to a camera. We mention some of the mathematical techniques borrowed from algebraic geometry, projective geometry, and homotopy theory that are required to solve these problems, list unsolved problems, and give some directions for future research. >

Dilated Networks for Photonic Switching

We present some novel architectures for rearrangeably nonblocking multistage photonic space switches implemented using arrays of Ti:LiNbO_{3} directional couplers. Multistage networks, studied mostly in the electronic domain, are obtained by minimizing the number of 2 × 2 elements needed to implement a switch. Unfortunately, straightforward extensions of these networks to the photonic domain show that the switch size has to be severely limited by the crosstalk in each of the Ti:LiNbO_{3} 2 \times 2 switching elements. Our networks, on the other hand, have a controllable (including almost zero) amount of crosstalk, low optical path loss, and an asymptotically optimal number of directional coupler switches for a given switch size. In addition, the switch has a simple control algorithm and its performance for light loading appears very promising. The switch is easily decomposable into smaller arrays of no more than two types, making it easy to partition the switch into chips. At the cost of a slight increase in crosstalk, the switch can be made single fault tolerant in terms of its ability to connect any input to any output.

Reconstruction filters in computer-graphics

Problems of signal processing arise in image synthesis because of transformations between continuous and discrete representations of 2D images. Aliasing introduced by sampling has received much attention in graphics, but reconstruction of samples into a continuous representation can also cause aliasing as well as other defects in image quality. The problem of designing a filter for use on images is discussed, and a new family of piecewise cubic filters are investigated as a practical demonstration. Two interesting cubic filters are found, one having good antialiasing properties and the other having good image-quality properties. It is also shown that reconstruction using derivative as well as amplitude values can greatly reduce aliasing.

Adaptive quantization of picture signals using spatial masking

Visual sensitivity of human observers decreases at and adjacent to large luminance changes: a fact well known to psychophysicists but not yet fully utilized for picture coding. In this paper we present a systematic investigation of these changes in visual sensitivity and apply it to adapt the quantizer of a predictive coder. The paper consists of three parts. In the first part, earlier psychophysical work in related areas relevant to picture coding is briefly reviewed. Certain simple measures of luminance activity (called the masking functions) are constructed. Subjective experiments which obtain fidelity measures related to the masking functions, using complex scenes (real-life pictures, head and shoulders view), are described. These relationships, called the visibility functions, express relationship between the relative amplitude accuracy required by the viewer and the masking functions. Perceptual, statistical, and contextual properties are inherent in them. Relationship between the visibility functions and some earlier measurements of psychovisual weighting functions are pointed out. In the second part, several adaptation strategies for the quantizer of a predictive DPCM coder are discussed. These include uniform as well as nonuniform quantizers with or without entropy constraints. These strategies are simulated on a computer, and the results are presented in the third part. The simulations indicate that, for the same picture quality, by using adaptive strategies, entropy reductions of about 30-50 percent are possible over nonadaptive techniques.

Design and implementation of a high-speed transport protocol

The design, analysis, and implementation of an end-to-end transport protocol that is capable of high throughput consistent with the evolving high-speed physical networks based on fiber-optic transmission lines and high-capacity switches are presented. Unlike current transport protocols in which changes in control/state information are exchanged between the two communicating entities only when some significant event occurs, this protocol exchanges relevant and full state information periodically and frequently. It is shown that this reduces the complexity of protocol processing by removing many of the procedures required to recover from network inadequacies such as bit errors, packet loss, and out-of-sequence packets and makes it more amenable to parallel processing. Also, to increase channel utilization in the presence of high-speed, long-latency networks and to support diagrams, and efficient implementation of the selective repeat method of error control is incorporated in the protocol. An implementation using a Motorola 68030-based multiprocessor as a front-end processor is described. The current implementation can comfortably handle 10-15 kpackets/s. >

Motion interpolation by optimal control

Motion Interpolation, which arises in many situations such as Keyframe Animation, is the synthesis of a sequence of images portraying continuous motion by interpolating between a set of keyframes. If the keyframes are specified by parameters of moving objects at several instants of time, (e.g., position, orientation, velocity) then the goal is to find their values at the intermediate instants of time. Previous approaches to this problem have been to construct these intermediate, or in-between, frames by interpolating each of the motion parameters independently. This often produces unnatural motion since the physics of the problem is not considered and each parameter is obtained independently. Our approach models the motion of objects and their environment by differential equations obtained from classical mechanics. In order to satisfy the constraints imposed by the keyframes we apply external control. We show how smooth and natural looking interpolations can be obtained by minimizing a combination of the control energy and the roughness of the trajectory of the objects in 3D-space. A general formulation is presented which allows several trade-offs between various parameters that control motion. Although optimal parameter values resulting in the best subjectively looking motion are not yet known, our simulations have produced smooth and natural motion that is subjectively better than that produced by other interpolation methods, such as the cubic splines.

Multicast Scheduling in Cellular Data Networks

Multicast is an efficient means of transmitting the same content to multiple receivers while minimizing network resource usage. Applications that can benefit from multicast such as multimedia streaming and download, are now being deployed over 3G wireless data networks. Existing multicast schemes transmit data at a fixed rate that can accommodate the farthest located users in a cell. However, users belonging to the same multicast group can have widely different channel conditions. Thus existing schemes are too conservative by limiting the throughput of users close to the base station. We propose two proportional fair multicast scheduling algorithms that can adapt to dynamic channel states in cellular data networks that use time division multiplexing: Inter-group Proportional Fairness (IPF) and multicast proportional fairness (MPF). These scheduling algorithms take into account (1) reported data rate requests from users which dynamically change to match their link states to the base station, and (2) the average received throughput of each user inside its cell. This information is used by the base station to select an appropriate data rate for each group. We prove that IPF and MPF achieve proportional fairness among groups and among all users in a group inside a cell respectively. Through extensive packet-level simulations, we demonstrate that these algorithms achieve good balance between throughput and fairness among users and groups.

Time-recursive deinterlacing for IDTV and pyramid coding

Most improved-definition television (IDTV) receivers use progressive scanning to reduce artifacts associated with interlacing (e.g. interline flicker, line crawl). Some novel techniques of motion compensated interpolation of the missing lines of interlaced monochrome and color sequences, reducing the artifacts associated with interlacing, and effectively increasing the vertical resolution of the image sequences are proposed. Time-recursive motion compensation prediction is introduced, in which all previously displayed history (not just the previous field) is used to predict the missing pixel values. The next future field is also used for the same purpose, by a lookahead scheme. Motion estimation is done using a quadtree-based segmented block-matching technique with half-pixel accuracy. To avoid artifacts and obtain full resolution in still regions, such as background, motion adaptation is also used. How to apply this algorithm to pyramid coding to achieve a better compression rate and compatibility with other lesser resolution standards is discussed. >

Dynamic maintenance and visualization of molecular surfaces

Molecular surface computations are often necessary in order to perform synthetic drug design. A critical step in this process is the computation and update of an exact boundary representation for the molecular surface (e.g. the Lee--Richards surface). In this paper we introduce efficient techniques for computing a molecular surface boundary representation as a set of non-uniform rational B-splines patches. This representation introduces for molecules the same geometric data structure used in the solid modeling community and enables immediate access to a wide range of modeling operations and techniques. Furthermore, this allows the use of any general solid modeling or visualization system as a molecular modeling interface. However, using such a representation in a molecular modeling environment raises several efficiency and update constraints, especially in a dynamic setting. For example, changes in the probe radius result in both geometric and topological changes to the set of patches. Our techniques provide the option of trading accuracy of the representation for the efficiency of the computation, while still tracking the changes in the set of patches. In particular, we discuss two main classes of dynamic updates: one that keeps the topology of the molecular configuration fixed, and a more complicated case were the topology may be updated continuously. In general the generated output surface is represented in a format that can be loaded into standard solid modeling systems. It can also be directly triangulated or rendered, possibly at different levels of resolution, by a standard graphics library such as OpenGL without any additional effort.