Pyramidal Architectures For Computer Vision

Computer vision deals with the problem of manipulating information contained in large quantities of sensory data, where raw data emerge from the transducing 6 7 sensors at rates between 10 to 10 pixels per second. Conventional general purpose computers are unable to achieve the computation rates required to op erate in real time or even in near real time, so massively parallel systems have been used since their conception in this important practical application area. The development of massively parallel computers was initially character ized by efforts to reach a speedup factor equal to the number of processing elements (linear scaling assumption). This behavior pattern can nearly be achieved only when there is a perfect match between the computational struc ture or data structure and the system architecture. The theory of hierarchical modular systems (HMSs) has shown that even a small number of hierarchical levels can sizably increase the effectiveness of very large systems. In fact, in the last decade several hierarchical architectures that support capabilities which can overcome performances gained with the assumption of linear scaling have been proposed. Of these architectures, the most commonly considered in com puter vision is the one based on a very large number of processing elements (PEs) embedded in a pyramidal structure. Pyramidal architectures supply the same image at different resolution lev els, thus ensuring the use of the most appropriate resolution for the operation, task, and image at hand.

This book contains the proceedings of the NATO Advanced Research Workshop held in Maratea (Italy), May 5-9, 1986 on Pyramidal Systems for Image Processing and Computer Vision. We had 40 participants from 11 countries playing an active part in the workshop and all the leaders of groups that have produced a prototype pyramid machine or a design for such a machine were present. Within the wide field of parallel architectures for image processing a new area was recently born and is growing healthily: the area of pyramidally structured multiprocessing systems. Essentially, the processors are arranged in planes (from a base to an apex) each one of which is generally a reduced (usually by a power of two) version of the plane underneath: these processors are horizontally interconnected (within a plane) and vertically connected with "fathers" (on top planes) and "children" on the plane below. This arrangement has a number of interesting features, all of which were amply discussed in our Workshop including the cellular array and hypercube versions of pyramids. A number of projects (in different parts of the world) are reported as well as some interesting applications in computer vision, tactile systems and numerical calculations.

Biological visual systems employ massively parallel processing to perform real-world visual tasks in real time. A key to this remarkable performance seems to be that biological systems construct representations of their visual image data at multiple scales. A Pyramid Framework for Early Vision describes a multiscale, or `pyramid', approach to vision, including its theoretical foundations, a set of pyramid-based modules for image processing, object detection, texture discrimination, contour detection and processing, feature detection and description, and motion detection and tracking. It also shows how these modules can be implemented very efficiently on hypercube-connected processor networks. A Pyramid Framework for Early Vision is intended for both students of vision and vision system designers; it provides a general approach to vision systems design as well as a set of robust, efficient vision modules.

Human performance in visual perception by far exceeds the performance of contemporary computer vision systems. While humans are able to perceive their environment almost instantly and reliably under a wide range of conditions, computer vision systems work well only under controlled conditions in limited domains. This book sets out to reproduce the robustness and speed of human perception by proposing a hierarchical neural network architecture for iterative image interpretation. The proposed architecture can be trained using unsupervised and supervised learning techniques. Applications of the proposed architecture are illustrated using small networks. Furthermore, several larger networks were trained to perform various nontrivial computer vision tasks.

Parallel Computer Vision

The exponential explosion of images and videos concerns everybody's common life, since this media is now present everywhere and in all human activities. Scientists, artists and engineers, in any field, need to be aware of the basic mechanisms that allow them to understand how images are essentially information carriers. Images bear a strong evocative power because their perception quickly brings into mind a number of related pictorial contents of past experiences and even of abstract concepts like pleasure, attraction or aversion. This book analyzes the visual hints, thanks to which images are generally interpreted, processed and exploited both by humans and computer programs. Comprehensive introductory text Introduces the reader to the large world of imagery on which many human activities are based, from politics to entertainment, from technical reports to artistic creations Provides a unified framework where both biological and artificial vision are discussed through visual cues, through the role of contexts and the available multi-channels to deliver information

Computer vision solutions used to be very specific and difficult to adapt to different or even unforeseen situations. The current development is calling for simple to use yet robust applications that could be employed in various situations. This trend requires the reassessment of some theoretical issues in computer vision. A better general understanding of vision processes, new insights and better theories are needed. The papers selected from the conference staged in Dagstuhl in 1996 to gather scientists from the West and the former eastern-block countries address these goals and cover such fields as 2D images (scale space, morphology, segmentation, neural networks, Hough transform, texture, pyramids), recovery of 3-D structure (shape from shading, optical flow, 3-D object recognition) and how vision is integrated into a larger task-driven framework (hand-eye calibration, navigation, perception-action cycle).

This book contains a selection of papers which were presented at the Vision Interface '92 Conference. It also includes several invited articles from prominent researchers in the field, suggesting future directions in Computer Vision.