• Title/Summary/Keyword: 마스크 처리

Search Result 382, Processing Time 0.024 seconds

Optical Implementation of Associative Memory Based on the Hopfield Model (Hopfield 모델에 기초한 연상 메모리의 광학적 구현)

  • 이재수;이승현;이우상;김은수
    • The Journal of Korean Institute of Communications and Information Sciences
    • /
    • v.14 no.5
    • /
    • pp.561-570
    • /
    • 1989
  • In this paper, we describe the theoretical analysis and optical implementation of real-time associative memory based on the modified Hopfield neural network model by using a commerical LCTV connected to computer graphic as the real-time memory mask and adding one mask line to the momory mask in order to optically obtain the time-varing thresholding values of the modified Hopfield model.

  • PDF

An Edge Detection Algorithm using Modified Mask in AWGN Environment (AWGN 환경에서 변형된 마스크를 이용한 에지 검출 알고리즘)

  • Lee, Chang-Young;Kim, Nam-Ho
    • Proceedings of the Korean Institute of Information and Commucation Sciences Conference
    • /
    • 2013.05a
    • /
    • pp.892-894
    • /
    • 2013
  • Edge has been utilized in various application fields with development of technique of digital image processing. In conventional edge detection methods, there are some methods using mask including Sobel, Prewitt, Roberts and Laplacian operator. Those methods are that implement is simple but generates errors of edge detection in images added AWGN(additive white Gaussian noise). Therefore, to compensate the defect of those methods, in this paper, an edge detection algorithm using modified mask is proposed, and it showed superior edge detection property in AWGN.

  • PDF

Post-Training with Hierarchical Masked Language Modeling (계층적 마스크 모델링을 이용한 언어 모델의 사후 학습)

  • Hyun-Kyu Jeon;Hyein Jung;Seoyeon Park;Bong-Su Kim
    • Annual Conference on Human and Language Technology
    • /
    • 2022.10a
    • /
    • pp.588-591
    • /
    • 2022
  • 최근 자연어 이해 및 생성에 있어서 사전학습 기반의 언어 모델이 널리 사용되고 있다. BERT, roBERTa 등의 모델이 있으며, 마스크 언어 모델링을 주요 과제로 하여 사전 학습을 한다. 하지만 MLM은 문법적인 정보를 활용하지 못하는 단점이 있다. 단 순히 무작위로 마스크를 씌우고 맞추기 때문이다. 따라서 본 연구에서는 입력 문장의 문법적 정보를 활용하는 방법을 소개하고, 이를 기반으로 사후 학습을 하여 그 효과를 확인해 본다. 공개된 사전학습 모델과 사후학습 모델을 한국어를 위한 벤치마크 데이터셋 KLUE에 대하여 조정학습하고 그 결과를 살펴본다.

  • PDF

A Study on the Effect of Fast Burn for Different Combustion Chamber Geometries of Gasoline Engine Using an Ion Current Method (이온전류법에 의한 가솔린엔진 연소실 형상별 급속연소효과 연구)

  • 강건용;서승우;정동수;장영준
    • Transactions of the Korean Society of Mechanical Engineers
    • /
    • v.17 no.6
    • /
    • pp.1633-1639
    • /
    • 1993
  • In spark ignition engine, EGR of lean mixture operation has advantage in emission, but disadvantages in power output and combustion flame propagation. Fast burn system is known to be a useful method to solve these disadvantages. This paper presents the characteristics of in-cylinder flow for different combustion chamber geometries, and the correlation between the in-cylinder flow and the combustion flame speed using an ion current method.

Design of Wearable IoT based Smart Mask (웨어러블 IoT기반 스마트 마스크 설계)

  • Park, Yonghyun;Jeong, SeongWoon;Jung, Kyung Kwon
    • Proceedings of the Korean Institute of Information and Commucation Sciences Conference
    • /
    • 2021.10a
    • /
    • pp.300-302
    • /
    • 2021
  • Usage of a face mask has become mandatory in many countries after the COVID-19. This paper described to develop a IoT based smart mask system for monitoring face mask. The system developed in this paper has two main units, a sensor module, and a smartphone application. The sensor module consists of four components: temperature and humidity sensor, a heart rate sensor, and a BLE chip. This components work as a unit to collect data and stream them through an I2C port over BLE to a connected mobile device. The smartphone application is an Android application developed for smart phones. It enables the Android device to communicate with the sensor to receive sensor data, process, store and display results.

  • PDF

The Robustness Wavelet Watermarking with Adaptive Weight MASK (적응 가중치 마스크 처리 기반 강인한 웨이브릿 워터마킹)

  • 정성록;김태효
    • Journal of the Institute of Convergence Signal Processing
    • /
    • v.4 no.2
    • /
    • pp.46-52
    • /
    • 2003
  • In this paper, the wavelet watermarking algorithm based on adaptive weight MASK processing as a watermark embedded-method for Copyright Protection of Digital contents is Proposed. Because watermark induce as a noise of original image, the watermark size should be limited for preventing quality losses and embedding watermark into images. Therefore, it should be preserve the best condition of the factors, robustness, capacity and visual quality. Tn order to solve this problem, we propose watermarking embedded method by applying adaptive weight MASK to the algorithm and optimize its efficiency. In that result, the watermarked images are improved about external attack. Specifically, correlation coefficient has over 0.8 on both modifications of brightness and contrast. Also, correlation coefficient of wavelet compression of embedded watermark last by over 0.65.

  • PDF

A Technique for Image Processing of Concrete Surface Cracks (콘크리트 표면 균열의 영상 처리 기법)

  • Kim Kwang-Baek;Cho Jae-Hyun;Ahn Sang-Ho
    • Journal of the Korea Institute of Information and Communication Engineering
    • /
    • v.9 no.7
    • /
    • pp.1575-1581
    • /
    • 2005
  • Recently, further study is being done on the affect of crack on concrete structure and many people have made every endeavor not to leave it unsettled but to minimize it by repair works. In this paper we propose the image processing method that do not remain manual but automatically process the length, the direction and e width of cracks on concrete surface. First, we calibrate light's affect from image by using closing operation, one of morphology methods that can extract the feature of oracle and we extract the edge of crack image by sobel mask. After it, crack image is binarized by iteration binarization. And we extract the edge of cracks using noise elimination method that use an average of adjacent pixels by 3${\times}$3 mask and Glassfire Labeling algorithm. on, in this paper we propose an image processing method which can automatically measure the length, the direction and the width of cracks using the extracted edges of cracks. The results of experiment showed that the proposed method works better on the extraction of concrete cracks. Also our method showed the possibility that inspector's decision is unnecessary.

An Adaptive Region-of-Interest Coding Based on EBCOT (EBCOT 기반의 적응적 관심영역 코딩)

  • Kang, Ki-Jun;Lee, Bu-Kwon;Seo, Yeong-Geon
    • Journal of Korea Multimedia Society
    • /
    • v.9 no.11
    • /
    • pp.1445-1454
    • /
    • 2006
  • To compress a specific part of an image with high quality or to transfer it, JPEG2000 standard offers an ROI(Region-of-Interest) image coding method. What is important in ROI coding is to process relative importance between ROI and background and to process ROI mask. We propose an adaptive ROI coding method supplemented the existing Implicit ROI coding and Modified implicit ROI coding to improve image quality and reduce ROI mask information. The proposed method is an EBCOT-based ROI coding that extracts ROI from the compressed bitstream, and gets the ROI mask information by classifying the codeblocks into 6 patterns. The information includes the pattern type(3bit) and the width(5bit) expressing the boundary between two regions for each codeblock. As a result, the method shows an excellent compression performance in ROI region as well as in the whole region of an image.

  • PDF

Salt and Pepper Noise Removal Algorithm based on Euclidean Distance Weight (유클리드 거리 가중치를 기반한 Salt and Pepper 잡음 제거 알고리즘)

  • Chung, Young-Su;Kim, Nam-Ho
    • Journal of the Korea Institute of Information and Communication Engineering
    • /
    • v.26 no.11
    • /
    • pp.1637-1643
    • /
    • 2022
  • In recent years, the demand for image-processing technology in digital marketing has increased due to the expansion and diversification of the digital market, such as video, security, and machine intelligence. Noise-processing is essential for image-correction and reconstruction, especially in the case of sensitive noises, such as in CT, MRI, X-ray, and scanners. The two main salt and pepper noises have been actively studied, but the details and edges are still unsatisfactory and tend to blur when there is a lot of noise. Therefore, this paper proposes an algorithm that applies a weight-based Euclidean distance equation to the partial mask and uses only the non-noisy pixels that are the most similar to the original as effective pixels. The proposed algorithm determines the type of filter based on the state of the internal pixels of the designed partial mask and the degree of mask deterioration, which results in superior noise cancellation even in highly damaged environments.

A Filter Algorithm based on Partial Mask and Lagrange Interpolation for Impulse Noise Removal (임펄스 잡음 제거를 위한 부분 마스크와 라그랑지 보간법에 기반한 필터 알고리즘)

  • Cheon, Bong-Won;Kim, Nam-Ho
    • Journal of the Korea Institute of Information and Communication Engineering
    • /
    • v.26 no.5
    • /
    • pp.675-681
    • /
    • 2022
  • Recently, with the development of IoT technology and AI, unmanned and automated in various fields, interest in video processing, which is the basis for automation such as object recognition and object classification, is increasing. Various studies have been conducted on noise removal in the video processing process, which has a significant impact on image quality and system accuracy and reliability, but there is a problem that it is difficult to restore images for areas with high impulse noise density. In this paper proposes a filter algorithm based on partial mask and Lagrange interpolation to restore the damaged area of impulse noise in the image. In the proposed algorithm, the filtering process was switched by comparing the filtering mask with the noise estimate and the purge weight was calculated based on the low frequency component and the high frequency component of the image to restore the image.