This paper seeks to describe work in progress towards the development of a very-highspeed (80-200 Mbits/sec) Local Area Network (LAN) for the support of a distributedimageprocessing system. The current practice in m...
详细信息
An image generation system by ray tracing was implemented on a distributedparallelprocessing system composed of loosely coupled multimicrocomputers. An improvement was made by extending the architecture from the sta...
详细信息
An image generation system by ray tracing was implemented on a distributedparallelprocessing system composed of loosely coupled multimicrocomputers. An improvement was made by extending the architecture from the star structure to the hierarchical tree structure. This improvement makes it possible for an animated image to be generated at high speed with shading which can represent the shape and texture of three-dimensional objects with high quality. This paper outlines the hierarchical tree-structured image generation system, discussing the parallelprocessing mechanisms of the system, such as data transfer and hierarchical load distribution schemes. Experiments were made to evaluate the performance. Problems in the tree-structured system are pointed out, and remedies are proposed (area redistribution, and area subdivision). Experiments are made to evaluate the effectiveness of these methods. Thus, it is demonstrated that the tree-structured system is useful in the image generation using ray tracing.
The feature extraction step is one of the most important in the recognition process of unknown patterns. As is well known, very fast feature extraction algorithms are required in a very large number of pattern recogni...
详细信息
ISBN:
(纸本)0892526920
The feature extraction step is one of the most important in the recognition process of unknown patterns. As is well known, very fast feature extraction algorithms are required in a very large number of pattern recognition fields. One of the newest methods for solving this problem very quickly is the use of parallel machines. In this work, a new type of Cellular Turing Acceptor is presented. Some of its applications for feature extraction from chain-codes are also reported.
The distributed Array Processor is a two dimensional interconnected array of 4096 processors, each with its own associated memory. Programming techniques on the DAP are often radically different from their serial coun...
详细信息
The distributed Array Processor is a two dimensional interconnected array of 4096 processors, each with its own associated memory. Programming techniques on the DAP are often radically different from their serial counterparts and two techniques that fit well on the DAP architecture are logical manipulation and recursive doubling. Both of these techniques are used in graphics and imageprocessing applications in areas as far apart as basic line drawing, and fourier transforms. Examples will be given of the type of performance and results available from these methods.
In this paper we describe some aspects of specifying, designing and evaluating a specialized machine, Romuald, for the capture, coding, and processing of video and scanning electron microscope (SEM) pictures. First we...
详细信息
The authors have presented nine different ways to propagate a seed value into a region of a binary image. They summarize the time complexities of these methods. An additional method recursive data-dependent scan has b...
详细信息
ISBN:
(纸本)0892524707
The authors have presented nine different ways to propagate a seed value into a region of a binary image. They summarize the time complexities of these methods. An additional method recursive data-dependent scan has been added because it is commonly used in practice;there a pushdown stack is used to implement a kind of backtracking search to perform a polygon-fill propagation. For convex, thick objects the fastest approach is parallel pyramidal using (4n**2/3) processors and the best algorithm for a distributedparallel system is distributed pyramidal.
This paper presents efficient methods of parallel-pipeline Walsh-Hadamard transformation for fast coding of decomposed pictures. The decomposition of the pictures is realized by the known regular decomposition methodo...
详细信息
This paper presents efficient methods of parallel-pipeline Walsh-Hadamard transformation for fast coding of decomposed pictures. The decomposition of the pictures is realized by the known regular decomposition methodology. The regular decomposition procedure converts the picture into a tree presentation and compresses the original amount of picture-data by rejecting non-informative areas of the picture. The Walsh-Hadamard transformation is performed on the regularly-decomposed picture as an averaging or smearing process to allow transmission which is more immune to channel errors. It also makes bandwidth reduction possible because, since the image energy which is usually uniformaly distributed in spatial domain now tends to be concentrated near the origin of the transform domain, some of the low-magnitude samples may be discarded.
During the last decade several parallel computers were developed, like ILLIAC iv, STARAN, CLIP 3 or DAP, which consist of many identical processors connected to each other in a very regular way. Such structures were i...
详细信息
During the last decade several parallel computers were developed, like ILLIAC iv, STARAN, CLIP 3 or DAP, which consist of many identical processors connected to each other in a very regular way. Such structures were investigated theoretically first by J. v. Neumann and are called ″cellular″ or ″iterative″ arrays. Meanwhile they are recognized as fast and efficient, because highly parallel, devices or models in order to formulate or to perform certain tasks of pattern recognition and transformations. The author knows three methods which are commonly used in order to design cellular algorithms: the simulation of waves, the simulation of particles and the filtering method. Advantages and disadvantages are outlined. A fourth method, called the ″local recognition method″ , is presented.
In recent years ″software engineering″ has emerged as a discipline of programming. It includes the conceptualization, design, implementation, testing and modification of software systems. Related issues are language...
详细信息
In recent years ″software engineering″ has emerged as a discipline of programming. It includes the conceptualization, design, implementation, testing and modification of software systems. Related issues are languages, standards, distributed and parallelprocessing, and total programming environments. The fact that more than 80 percent of system development costs were in software rather than hardware helps one appreciate the importance of any effort to understand and enhance the software production process. Many pattern recognition projects involve fairly large software efforts. It makes sense not only for researchers to make use of the latest software tools and methodologies but also to anticipate future changes.
During the last decade several parallel computers were developed, like ILLIAC iv, STARAN, CLIP 3 or DAP, which consist of many identical processors connected to each other in a very regular way. Such structures were i...
详细信息
During the last decade several parallel computers were developed, like ILLIAC iv, STARAN, CLIP 3 or DAP, which consist of many identical processors connected to each other in a very regular way. Such structures were investigated theoretically first by J. v. Neumann and are called ″cellular″ or ″iterative″ arrays. Meanwhile they are recognized as fast and efficient, because highly parallel, devices or models in order to formulate or to perform certain tasks of pattern recognition and transformations. The author knows three methods which are commonly used in order to design cellular algorithms: the simulation of waves, the simulation of particles and the filtering method. Advantages and disadvantages are outlined. A fourth method, called the ″local recognition method″ , is presented.
暂无评论