Hitoshi Matsushima

A multiprocessor system utilizing enhanced DSPs for image processing

A general-purpose image processor (GPIP) consisting of 64 digital signal processors (DSPs) in a 0.31-m/sup 3/ box is proposed to perform a wide range of image processing tasks. A high-speed DSP called DSP-i has been especially developed for this purpose. It has a highly parallel architecture with a two-level instruction hierarchy, multibank cache, and multiprocessor interface. The DSP-i machine cycle is 50 ns. A novel ring shift register bus architecture offers a flexible structure and an efficient data-exchange method for the system. Along with four proposed operation modes, it cuts the multiprocessing overhead to as little as 20%. The performance of the GPIP is 1000 MOPS (million operations per second).<<ETX>>

A 50ns DSP with parallel processing architecture

This report will cover a programmable DSP that avoids I/O bottlenecks through a two-level hierarchy of instructions. The IC contains 430K transistors on a 149mm2die.

Contour representation of binary images using run-type direction codes

Abstract.This paper proposes a fast and high-definition contour-tracing algorithm for digitized binary images. Only the coordinates of white-to-black and black-to-white transition points on each scan line are used as data for the contour tracing. The resulting description of the contour is the combined set of ten newly developed run-type direction codes. The code is determined automatically by an automaton while the horizontal coordinates of two transition points in the two adjacent scan lines are compared. The processing speed of this algorithm depends on the number of transition points of an image, and, thus, it has the advantage over conventional algorithms, where the speed usually depends on the number of pixels. A high-definition-type contour description is obtained and complete restoration of the original image is possible.

Texture discrimination using self-organized multiresolution filtering

In texture discrimination the basic image processes are considered to choose effective resolution, to determine representative features and to unify subregions having the same representative features. To realize these processes, this paper presents an improved method for discriminating higher-order features by using a self-organized multiresolutional filter. First, we make a local transformation function with the maximum discrimination ability by a limited-size neighborhood operation. Next, extending the method to the multiresolutional case, we let neighbor pixels of resolution which is optimal for discrimination be chosen automatically. Moreover, we realize unification of texture regions by connecting features represented by these pixel values and by obtaining macro-features. In this paper we also describe examples in which this method is applied to extract title characters from printed pages and to extract a specific part from a map image, and show that this method using higher-order features possess sufficient discrimination ability for real texture images.

Study of impingement cooling of heat sinks for LSI packages with longitudinal fins

This paper describes an experimental and a semi-empirical study on the impingement cooling characteristics of heat sinks with longitudinal fins of a type suitable for LSI packages. The experiments were performed with a variety of different fins. To enhance impingement cooling, one long rectangular inlet orifice (slit) over the center of the heat sink was found to offer the best structure. The optimum orifice width is about 1/6 of the base width of the heat sink. The thermal resistance at a fixed volumetric flow rate and orifice width varies little with size of the gap between the fin tops and inlet orifice, but gaps near 2 mm slightly lower the resistance. Correlations are proposed between the thermal resistance of the heat sink and the geometry of the longitudinal fins. The accuracy of the predicted thermal resistances was found to be within ±25% of the experimental data.

An Array Processor for Image Processing

Some early developments in pattern processing are reviewed in Unger (1958), McCormick (1963), and Murtha (1966). Some practical implementations are described by Golay (1969) and by Kruse (1976). For certain industrial and medical applications special processors for image analysis are in practical use such as in computed tomog-graphy (CT). Also, large scale pattern processors have become available in special fields.

Optimization of heat sink geometries for impingement air‐cooling of LSI packages

This paper describes the use of our previous studys prediction procedures for calculating thermal resistance and pressure drop. The procedures are used in the optimization of heat sink geometries for impingement air-cooling of LSI packages. Two types of heat sinks are considered: ones with longitudinal fins and ones with pin fins. We optimized the heat sink geometries by evaluating 16 parameters simultaneously. The parameters included fin thickness, spacing, and height. For the longitudinal fins, the optimal fin thicknesses were found to be between 0.12 and 0.15 mm, depending on which of the four types of fans were used. For pin fins, the optimal pin diameters were between 0.39 and 0.40 mm. Under constant pumping power, the optimal thermal resistance of the longitudinal fins was about 60% that of the pin fins. For both types of heat sinks, the optimal thermal resistance for four off-the-shelf fans was only slightly (maximum about 1%) higher than the theoretical optimum for the same pumping power. When manufacturing cost performance is considered, the most economical fin thickness and diameter are about 5 to 10 times higher than the optimal values calculated without respect for manufacturing costs. These values almost correspond to the actual limits of extrusion and press heat-sink manufacturing processes.

Knowledge-based multi-media database management system and its application to map information systems

A structure of a multimedia database management system that enables inferential retrieval and automatic construction is proposed. It consists of three parts, namely, a knowledge base at the top, a relational manager in the middle, and a media-oriented manager at the bottom. An automatic recognition and data input method for map figure data and an automatic database construction method and renewal capability are introduced. The effectiveness of these methods has been confirmed by applying them to a prototype map-based system for utility management.<<ETX>>

Numerical Simulation of Impingement Air Cooling From LSI Packages With Large Plate Fins by the Penalty Finite Element Method

Steady state three-dimensional incompressible flow analysis based on the two-equation model of turbulence is performed. The finite element method employing the penalty function formulation by Hughes et. al (1979) is used. To reduce the computational time and computer storage, the conjugate gradient method is applied to solve algebraic equations. In applying the conjugate gradient method, the equations are preconditioned so as to arrive at converged solutions effectively.

Prediction algorithm of thermal resistance for impingement cooling of heat sinks for LSI packages with pin-fin arrays

This paper is a semi-empirical report on an algorithm for the prediction of thermal resistance for impingement cooling of pin-fin heat sinks for LSI packages when the inlet orifice is relatively large and is located over the center of the sink. We present a physical model suitable for these types of heat sinks, based on flow visualization results. The model divides the flow region into five parts: I) the top surfaces of the fins where they are directly under the inlet orifice, II) the portions of the vertical surfaces of the pin-fin cylinders, where those surfaces are directly below the inlet port, III) the surface of the base to which the fins are attached, excluding the areas occupied by the feet of the fins themselves, IV) the portions of the vertical surfaces of the fin-cylinders excluding those portions of the surfaces that are directly below the inlet port (complementary to region II), V) the portions of the top surfaces of the pins, excluding those portions directly below the inlet port (complementary to region I). We predicted thermal resistance values for heat sinks with pin-fin arrays, for a variety of orifice diameters, gaps, pin-fin diameters, and heights, and number of fins. These values agreed with experimental data within ±30%.