• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.11a
• /
• pp.799-802
• /
• 2017

We attempt to combines concepts of Morphological Operator(MO) and Convolutional Neural Networks(CNN) to improve image-to-image translation. To do this, we propose an operation that approximates morphological operations. Also we propose S-Convolution, an operation that extends the operation to use multiple filters like CNN. The experiment result shows that it can learn MO with big filter using multiple S-convolution layer of small filter. To validate effectiveness of the proposed layer in image-to-image translation we experiment with GAN with S-convolution applied. The result showed that GAN with S-convolution can achieve distinct result from that of GAN with CNN.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.49-51
• /
• 2005

시간의 흐름에 따라 그 위치가 빈번히 변화하는 이동 객체의 특성으로 인해 실시간으로 증가하는 이동객체의 연산 정보를 효과적으로 관리할 수 있는 효율적인 연산 기법이 요구된다. 따라서 본 논문에서는 이동 객체의 시공간 연산을 위해 기존에 제안되었던 HR 트리의 성능을 개선시킬 수 있는 확장된 HR-트리(Extended HR-Tree : EHR-Tree)를 제안한다. 기존의 HR 트리는 삽입, 삭제, 갱신과 같은 연산이 빈번한 경우에 단말 노드 및 비단말 노드를 새로 생성해야 함으로 인하여 성능이 떨어지고, 공간의 낭비가 있게 된다. 이 문제를 해결하기 위해 본 논문에서 제안하는 EHR-트리는 연산이 발생할 때마다 새로운 HR-트리를 생성하는 것이 아니라 시간 구간을 두어서 새로 발생한 연산이 같은 시간 구간에 있을 경우 그 단위시간에 생성된 HR-트리에 그대로 삽입, 삭제, 갱신과 같은 연산을 수행하게 된다. 따라서 기존 HR-트리에서 단말 노드 및 비단말 노드를 새로 생성해야함으로 발생되던 많은 저장 공간 요구를 감소시킴으로, 즉 전체 연산 크기를 작게 하여 디스크 I/O수를 감소시킴으로써, 시공간 질의의 처리 속도를 향상시켜 효율적인 질의가 가능하도록 한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2004.05a
• /
• pp.23-26
• /
• 2004

최근 인터넷 및 이동 통신기기의 사용이 급증하면서 각종 데이터에 대한 사용자들의 검색 요청은 빠른 응답 시간을 요구하는 경우가 늘어나게 되었다. 이를 충족시키기 위하여 주기억 상주 데이터베이스 관리 시스템들이 등장하게 되었고 또한 대량의 데이터들에 대한 색인 구조와 색인에 대한 접근 충돌을 제어하면서도 빠른 응답 시간을 보장하는 색인 동시성 제어 기법들에 관한 연구가 활발히 진행되어 왔다. 현재 대부분의 주기억 상주 데이터베이스 관리 시스템들은 색인에 대한 동시성 제어 기법으로 잠금 기반의 동시성 제어 기법들을 많이 사용하고 있다. 그러나 잠금 기반의 동시성 제어 기법들은 검색 연산을 포함한 모든 연산에 대하여 접근하려고 하는 노드에 잠금을 거는 것을 요구하기 때문에 잠금이 걸려있는 노드를 접근 하려는 연산은 잠금이 풀리기를 기다려야만 한다. 따라서 잠금 기반의 색인 동시성 제어 기법들은 동시성에 제약이 생겨 검색 요청에 대한 응답시간을 지연시킨다. 본 논문에서는 $B^+$-Tree에서의 잠금 없는 검색 연산을 위하여 엔트리간 연결 리스트 기반의 동시성 제어 기법을 제안한다. 본 논문에서 제안하는 기법은 주기억 상주 데이터베이스 관리 시스템 환경에서 검색 연산이 아무런 잠금없이 수행되는 것을 보장한다. 특히 본 논문에서 제안하는 기법은 삽입, 삭제 연산이 수행 중인 노드에서의 잠금없는 검색 연산의 수행이 가능하기 때문에 잠금 기반의 동시성 제어 기법들 보다 빠르고 예측 가능한 응답시간을 보장한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.11a
• /
• pp.1506-1509
• /
• 2013

본 논문에서는 고성능 HEVC(High Efficiency Video Coding) 복호기 설계를 위한 효율적인 in-loop filter의 하드웨어 구조 설계에 대해 기술한다. in-loop filter는 deblocking filter와 SAO로 구성되며, 블록 단위 영상 압축 및 양자화 등에서 발생하는 정보의 손실을 보상하는 기술이다. 하지만 HEVC는 $64{\times}64$ 블록 크기까지 화소 단위 연산을 수행하기 때문에 높은 연산시간 및 연산량이 요구된다. 따라서 본 논문에서 제안하는 in-loop filter의 deblocking filter 모듈과 SAO 모듈은 최소 연산 단위인 $8{\times}8$ 블록 연산기로 구성하여 하드웨어 면적을 최소화하였다. 또한 SAO에서는 $8{\times}8$ 블록의 연산 결과를 내부레지스터에 저장하는 구조로 $64{\times}64$ 블록 크기를 지원하도록 설계하여 연산시간 및 연산량을 최소화 하였다. 제안하는 하드웨어 구조는 Verilog HDL로 설계하였으며, TSMC 칩 공정 180nm 셀 라이브러리로 합성한 결과 동작 주파수는 270MHz이고, 전체 게이트 수는 48.9k이다.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.38 no.12
• /
• pp.44-55
• /
• 2001

MAC(Multiply and ACcumulate) is the core operation of multimedia data processing. Because MAC units implemented on traditional DSP units or embedded processors have latency of three cycles and cannot operate on multiple data simultaneously, then, performances are seriously limited. Many high end general purpose microprocessors have SIMD MAC unit as a functional unit. But these high end MAC units must support pipeline structure for various operation modes and high clock frequency, which makes control logic complex and increases chip area. In this paper, a 64bit SIMD MAC unit for embedded processors is designed. It is implemented to have a latency of one clock cycle to remove pipeline control logics and a minimal area overhead for SIMD support is added to existing Booth multipliers.

• Journal of KIISE
• /
• v.43 no.5
• /
• pp.541-552
• /
• 2016

This paper introduces the new notions of conjunctive and complement choices in process algebra, which reduce both process and system complexities significantly for distributed mobile real-time system during specification and analysis phases. The complement choice implies that two processes make cohesive choices for their synchronous partners at their own choice operations. The conjunctive choice implies choice dependency among consecutive choice operations in a process. The conjunctive choice reduces process complexity exponentially by the degree of the consecutive choice operations. The complement choice also reduces system complexity exponentially by the degree of the synchronous choice operations. Consequently, the reduction method makes the specification and analysis of the systems much easier since the complexity is reduced significantly. This notion is implemented in a process algebra, called ${\delta}$-Calculus. The efficiency and effectiveness are demonstrated with an example in a tool for the algebra, called SAVE, which is developed on ADOxx platform.

• Journal of KIISE
• /
• v.44 no.9
• /
• pp.871-877
• /
• 2017

A vast amount of research has been carried out for executing compute-intensive applications on resource-constrained mobile devices. Computation offloading is a method in which heavy computations are dynamically migrated from a mobile device to a server, exploiting the powerful hardware of the server to perform complex computations. An important issue for offloading is the complexity of reconciling the execution state of applications between the server and the client. To address this issue, snapshot-based offloading has recently been proposed, which utilizes the snapshot of a web app as the portable description of the execution state. However, for web applications using the HTML5 canvas, snapshot-based offloading does not function correctly, because the snapshot cannot capture the state of the canvas. In this paper, we propose a code generation technique to save the canvas state as part of a snapshot, so that the snapshot-based offloading can be applied to web applications using the canvas.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.6
• /
• pp.567-572
• /
• 2019

In this paper, we propose an algorithm to perform convolution, pooling, and ReLU operations in CNN using binary image and binary kernel. It decomposes 256 gray-scale images into 8 bit planes and uses a binary kernel consisting of -1 and 1. The convolution operation of binary image and binary kernel is performed by addition and subtraction. Logically, it is a binary operation algorithm using the XNOR and comparator. ReLU and pooling operations are performed by using XNOR and OR logic operations, respectively. Through the experiments to verify the usefulness of the proposed algorithm, We confirm that the CNN operation can be performed by converting it to binary logic operation. It is an algorithm that can implement deep running even in a system with weak computing power. It can be applied to a variety of embedded systems such as smart phones, intelligent CCTV, IoT system, and autonomous car.

• The Journal of Engineering Research
• /
• v.2 no.1
• /
• pp.31-37
• /
• 1997

This paper describes the design and implementation of 8 bit ALU graphic simulator which helps students who study the structure and operation course of general ALU. ALU of this paper consists of three parts, arithmetic circuit, logic circuit, and shifter. Each of them performs as follows. Arithmetic circuit performs arithmetic operation such as addition, subtraction, 1 increment, 1 decrement, 2's complement, logic circuit performs logic operation such as OR, AND, XOR, NOT, and shifter performs shift operation and transfers the result of circuits of arithmetic, logic to data bus. The instructions which relate to these basic ALU functions was selected from Z80 instructions and ALU circuit was designed with those instructions and this designed ALU circuit was implemented on graphic screen. And all state of this data operation course in ALU was showed by bit and logic gate unit.

• The Journal of the Petrological Society of Korea
• /
• v.6 no.1
• /
• pp.45-51
• /
• 1997

Tectonic movements in the Mesozoic were significant events to structural evolution in East China, so far as to West Pacific area. Typical Mesozoic structures were formed and outcropped in Yanshanian mountain area in which Yanshanian movement was named. It is generally considered that the most of outcropping structures in this area were formed in Yanshanian movement. But general studies indicated recently that more than half of the folds were formed in Yanshanian movement. But general studies indicated recently that more than half of the folds were formed and most of fault were in great reverse activity during Indosinian movement in Late-Triassic epoch. The tectonic dynamics setting of Indosinian move ment is a N-S compressive stress system originated by northward movement of Sino-Korean massif and its collison with Xingan-Mongolia fold zone. A series of closed folds (nearly E-W axial trace)and some overturned folds were formed in Indosinian movement and incoaxially superposed by Yanshanian deformation, Faulting characteristcs in the area were thrust faulting caused by compressive stress in Indosinian movement, some of which appear to be positive structural inversion, and oblique-thrust caused by compressive-shear in Yanshanian movement.