The Computation of Color

This thesis takes an interdisciplinary approach to the study of color vision, focussing on the phenomenon of color constancy formulated as a computational problem. The primary contributions of the thesis are (1) the demonstration of a formal framework for lightness algorithms; (2) the derivation of a new lightness algorithm based on regularization theory; (3) the synthesis of an adaptive lightness algorithm using "learning" techniques; (4) the development of an image segmentation algorithm that uses luminance and color information to mark material boundaries; and (5) an experimental investigation into the cues that human observers use to judge the color of the illuminant. Other computational approaches to color are reviewed and some of their links to psychophysics and physiology are explored.

ACTORS: A Model of Concurrent Computation in Distributed Systems

A foundational model of concurrency is developed in this thesis. We examine issues in the design of parallel systems and show why the actor model is suitable for exploiting large-scale parallelism. Concurrency in actors is constrained only by the availability of hardware resources and by the logical dependence inherent in the computation. Unlike dataflow and functional programming, however, actors are dynamically reconfigurable and can model shared resources with changing local state. Concurrency is spawned in actors using asynchronous message-passing, pipelining, and the dynamic creation of actors. This thesis deals with some central issues in distributed computing. Specifically, problems of divergence and deadlock are addressed. For example, actors permit dynamic deadlock detection and removal. The problem of divergence is contained because independent transactions can execute concurrently and potentially infinite processes are nevertheless available for interaction.

Dataflow Computation for the J-Machine

The dataflow model of computation exposes and exploits parallelism in programs without requiring programmer annotation; however, instruction- level dataflow is too fine-grained to be efficient on general-purpose processors. A popular solution is to develop a "hybrid'' model of computation where regions of dataflow graphs are combined into sequential blocks of code. I have implemented such a system to allow the J-Machine to run Id programs, leaving exposed a high amount of parallelism --- such as among loop iterations. I describe this system and provide an analysis of its strengths and weaknesses and those of the J-Machine, along with ideas for improvement.

Reconfigurable Architectures for General-Purpose Computing

General-purpose computing devices allow us to (1) customize computation after fabrication and (2) conserve area by reusing expensive active circuitry for different functions in time. We define RP-space, a restricted domain of the general-purpose architectural space focussed on reconfigurable computing architectures. Two dominant features differentiate reconfigurable from special-purpose architectures and account for most of the area overhead associated with RP devices: (1) instructions which tell the device how to behave, and (2) flexible interconnect which supports task dependent dataflow between operations. We can characterize RP-space by the allocation and structure of these resources and compare the efficiencies of architectural points across broad application characteristics. Conventional FPGAs fall at one extreme end of this space and their efficiency ranges over two orders of magnitude across the space of application characteristics. Understanding RP-space and its consequences allows us to pick the best architecture for a task and to search for more robust design points in the space. Our DPGA, a fine- grained computing device which adds small, on-chip instruction memories to FPGAs is one such design point. For typical logic applications and finite- state machines, a DPGA can implement tasks in one-third the area of a traditional FPGA. TSFPGA, a variant of the DPGA which focuses on heavily time-switched interconnect, achieves circuit densities close to the DPGA, while reducing typical physical mapping times from hours to seconds. Rigid, fabrication-time organization of instruction resources significantly narrows the range of efficiency for conventional architectures. To avoid this performance brittleness, we developed MATRIX, the first architecture to defer the binding of instruction resources until run-time, allowing the application to organize resources according to its needs. Our focus MATRIX design point is based on an array of 8-bit ALU and register-file building blocks interconnected via a byte-wide network. With today's silicon, a single chip MATRIX array can deliver over 10 Gop/s (8-bit ops). On sample image processing tasks, we show that MATRIX yields 10-20x the computational density of conventional processors. Understanding the cost structure of RP-space helps us identify these intermediate architectural points and may provide useful insight more broadly in guiding our continual search for robust and efficient general-purpose computing structures.

A Statistical Image-Based Shape Model for Visual Hull Reconstruction and 3D Structure Inference

We present a statistical image-based shape + structure model for Bayesian visual hull reconstruction and 3D structure inference. The 3D shape of a class of objects is represented by sets of contours from silhouette views simultaneously observed from multiple calibrated cameras. Bayesian reconstructions of new shapes are then estimated using a prior density constructed with a mixture model and probabilistic principal components analysis. We show how the use of a class-specific prior in a visual hull reconstruction can reduce the effect of segmentation errors from the silhouette extraction process. The proposed method is applied to a data set of pedestrian images, and improvements in the approximate 3D models under various noise conditions are shown. We further augment the shape model to incorporate structural features of interest; unknown structural parameters for a novel set of contours are then inferred via the Bayesian reconstruction process. Model matching and parameter inference are done entirely in the image domain and require no explicit 3D construction. Our shape model enables accurate estimation of structure despite segmentation errors or missing views in the input silhouettes, and works even with only a single input view. Using a data set of thousands of pedestrian images generated from a synthetic model, we can accurately infer the 3D locations of 19 joints on the body based on observed silhouette contours from real images.

Image-Based View Synthesis

We present a new method for rendering novel images of flexible 3D objects from a small number of example images in correspondence. The strength of the method is the ability to synthesize images whose viewing position is significantly far away from the viewing cone of the example images ("view extrapolation"), yet without ever modeling the 3D structure of the scene. The method relies on synthesizing a chain of "trilinear tensors" that governs the warping function from the example images to the novel image, together with a multi-dimensional interpolation function that synthesizes the non-rigid motions of the viewed object from the virtual camera position. We show that two closely spaced example images alone are sufficient in practice to synthesize a significant viewing cone, thus demonstrating the ability of representing an object by a relatively small number of model images --- for the purpose of cheap and fast viewers that can run on standard hardware.

Image Based Rendering Using Algebraic Techniques

This paper presents an image-based rendering system using algebraic relations between different views of an object. The system uses pictures of an object taken from known positions. Given three such images it can generate "virtual'' ones as the object would look from any position near the ones that the two input images were taken from. The extrapolation from the example images can be up to about 60 degrees of rotation. The system is based on the trilinear constraints that bind any three view so fan object. As a side result, we propose two new methods for camera calibration. We developed and used one of them. We implemented the system and tested it on real images of objects and faces. We also show experimentally that even when only two images taken from unknown positions are given, the system can be used to render the object from other view points as long as we have a good estimate of the internal parameters of the camera used and we are able to find good correspondence between the example images. In addition, we present the relation between these algebraic constraints and a factorization method for shape and motion estimation. As a result we propose a method for motion estimation in the special case of orthographic projection.

Model-Based Matching by Linear Combinations of Prototypes

We describe a method for modeling object classes (such as faces) using 2D example images and an algorithm for matching a model to a novel image. The object class models are "learned'' from example images that we call prototypes. In addition to the images, the pixelwise correspondences between a reference prototype and each of the other prototypes must also be provided. Thus a model consists of a linear combination of prototypical shapes and textures. A stochastic gradient descent algorithm is used to match a model to a novel image by minimizing the error between the model and the novel image. Example models are shown as well as example matches to novel images. The robustness of the matching algorithm is also evaluated. The technique can be used for a number of applications including the computation of correspondence between novel images of a certain known class, object recognition, image synthesis and image compression.

Example Based Image Analysis and Synthesis

Image analysis and graphics synthesis can be achieved with learning techniques using directly image examples without physically-based, 3D models. In our technique: -- the mapping from novel images to a vector of "pose" and "expression" parameters can be learned from a small set of example images using a function approximation technique that we call an analysis network; -- the inverse mapping from input "pose" and "expression" parameters to output images can be synthesized from a small set of example images and used to produce new images using a similar synthesis network. The techniques described here have several applications in computer graphics, special effects, interactive multimedia and very low bandwidth teleconferencing.

Exploring Object Perception with Random Image Structure Evolution

We have developed a technique called RISE (Random Image Structure Evolution), by which one may systematically sample continuous paths in a high-dimensional image space. A basic RISE sequence depicts the evolution of an object's image from a random field, along with the reverse sequence which depicts the transformation of this image back into randomness. The processing steps are designed to ensure that important low-level image attributes such as the frequency spectrum and luminance are held constant throughout a RISE sequence. Experiments based on the RISE paradigm can be used to address some key open issues in object perception. These include determining the neural substrates underlying object perception, the role of prior knowledge and expectation in object perception, and the developmental changes in object perception skills from infancy to adulthood.

People Recognition in Image Sequences by Supervised Learning

We describe a system that learns from examples to recognize people in images taken indoors. Images of people are represented by color-based and shape-based features. Recognition is carried out through combinations of Support Vector Machine classifiers (SVMs). Different types of multiclass strategies based on SVMs are explored and compared to k-Nearest Neighbors classifiers (kNNs). The system works in real time and shows high performance rates for people recognition throughout one day.

Perceptually-based Comparison of Image Similarity Metrics

The image comparison operation ??sessing how well one image matches another ??rms a critical component of many image analysis systems and models of human visual processing. Two norms used commonly for this purpose are L1 and L2, which are specific instances of the Minkowski metric. However, there is often not a principled reason for selecting one norm over the other. One way to address this problem is by examining whether one metric better captures the perceptual notion of image similarity than the other. With this goal, we examined perceptual preferences for images retrieved on the basis of the L1 versus the L2 norm. These images were either small fragments without recognizable content, or larger patterns with recognizable content created via vector quantization. In both conditions the subjects showed a consistent preference for images matched using the L1 metric. These results suggest that, in the domain of natural images of the kind we have used, the L1 metric may better capture human notions of image similarity.

Dissociated Dipoles: Image representation via non-local comparisons

A fundamental question in visual neuroscience is how to represent image structure. The most common representational schemes rely on differential operators that compare adjacent image regions. While well-suited to encoding local relationships, such operators have significant drawbacks. Specifically, each filter's span is confounded with the size of its sub-fields, making it difficult to compare small regions across large distances. We find that such long-distance comparisons are more tolerant to common image transformations than purely local ones, suggesting they may provide a useful vocabulary for image encoding. . We introduce the "Dissociated Dipole," or "Sticks" operator, for encoding non-local image relationships. This operator de-couples filter span from sub-field size, enabling parametric movement between edge and region-based representation modes. We report on the perceptual plausibility of the operator, and the computational advantages of non-local encoding. Our results suggest that non-local encoding may be an effective scheme for representing image structure.

Validating Co-Training Models for Web Image Classification

Co-training is a semi-supervised learning method that is designed to take advantage of the redundancy that is present when the object to be identified has multiple descriptions. Co-training is known to work well when the multiple descriptions are conditional independent given the class of the object. The presence of multiple descriptions of objects in the form of text, images, audio and video in multimedia applications appears to provide redundancy in the form that may be suitable for co-training. In this paper, we investigate the suitability of utilizing text and image data from the Web for co-training. We perform measurements to find indications of conditional independence in the texts and images obtained from the Web. Our measurements suggest that conditional independence is likely to be present in the data. Our experiments, within a relevance feedback framework to test whether a method that exploits the conditional independence outperforms methods that do not, also indicate that better performance can indeed be obtained by designing algorithms that exploit this form of the redundancy when it is present.