• 한국콘텐츠학회논문지
• /
• 제12권4호
• /
• pp.31-39
• /
• 2012

분산 비디오 압축 기술은 초경량 비디오 압축 기술로써 많은 주목을 받고 있으며, 대표적인 기법은 피드백 채널을 이용하여 우수한 부호화 성능을 유지한다. 그러나 이로 인해 복호화기의 복잡도를 증대시키고 매우 많은 반복적인 연산에 의한 큰 복호화 지연을 요구하기 때문에 실시간 구현에 제한이 되고 있으며, 이를 개선하기 위한 연구가 필요하다. 이를 위해, 본 논문에서는 화소 영역 위너-지브 비디오 복호화 기법에서 각 비트 플레인 내에 위치한 LDPCA 프레임간의 시간적 상관성, 공간적 상관성 그리고 시공간적 상관성 등을 고려한 패리티 비트 요구량에 대한 예측 방법을 제시하고 고속 분산 비디오 복호화기법에 적용하여 성능을 비교한다. 모의실험을 통해, 움직임이 큰 영상과 움직임이 적은 영상에 대해 각각 시공간적 상관성과 시간적 상관성을 이용한 방식이 우수한 특성을 보이며, 이는 분산 비디오 부호화 기법의 다양한 응용 환경에 따른 효과적인 패리티 요구량 예측기법을 찾는데 효과적으로 사용될 수 있을 것이다.

• Journal of information and communication convergence engineering
• /
• 제9권2호
• /
• pp.201-206
• /
• 2011

FRUC (Frame Rate Up-Conversion) technique needs an effective frame interpolation algorithm using motion information between adjacent neighboring frames. In order to have good visual qualities in the interpolated frames, it is necessary to develop an effective detection and interpolation algorithms for occluded regions. For this aim, this paper proposes an effective occluded region detection algorithm through the adaptive forward and backward motion searches and also by introducing the minimum value of normalized cross-correlation coefficient (NCCC). That is, the proposed scheme looks for the location with the minimum sum of absolute differences (SAD) and this value is compared to that of the location with the maximum value of NCCC based on the statistics of those relations. And, these results are compared with the size of motion vector and then the proposed algorithm decides whether the given block is the occluded region or not. Furthermore, once the occluded regions are classified, then this paper proposes an adaptive interpolation algorithm for occluded regions, which still exist in the merged frame, by using the neighboring pixel information and the available data in the occluded block. Computer simulations show that the proposed algorithm can effectively classify the occluded region, compared to the conventional SAD-based method and the performance of the proposed interpolation algorithm has better PSNR than the conventional algorithms.

• 한국통신학회논문지
• /
• 제28권2C호
• /
• pp.100-109
• /
• 2003

가변길이 나무구조 벡터양자화기(VLTSVQ : variable-length tree-structured vector quantizer)를 기반으로 하는 영상 부호화 방식인 차분 인덱스(DI : Differential index) 할당기법을 제안하였다. 각 소스벡터는 VLTSVQ의 단말 노드로 양자화가 되어지고, 각 단말 노드는 유일한 이진 벡터로 표현된다. 제안한 방법은 영상의 화질은 유지하면서 압축률을 개선하기 위해 이웃하는 영상 블록들간의 상관성을 이용하였다. 모의실험을 통하여 제안한 방법이 기존 방법들에 비해 매우 높은 압축률을 보였으며, 화소간의 상관성이 높은 영상에 대해서는 63.2%의 발생 비트율 감소를 확인하였다. 그리고 영상내의 이웃블록들간 상관성이 커질수록 더 큰 비트율이 감소됨을 보였다. 제안한 부호화 기법은 일반적으로 이웃된 화소들간에 높은 상관성을 가진 MR(magnetic resonance)영상 부호화에 효율적으로 사용될 수 있다.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2007년도 Proceedings of ISRS 2007
• /
• pp.580-583
• /
• 2007

Various matching methods have been introduced by investigators to improve digital elevation model (DEM) accuracy of satellite imagery. This study proposed an area-based matching method according to land cover property using correlation coefficient of pixel brightness value between the two images for DEM generation from IKONOS stereo imagery. For this, matching line (where "matching line" implies straight line that is approximated to complex nonlinear epipolar geometry) is established by exterior orientation parameters to minimize search area. The matching is carried out based on this line. Land cover classes are divided off into water, urban land, forest and agricultural land. Matching size is selected using a correlation-coefficient image in the four areas. The selected sizes are $81{\times}81$ pixels window, $21{\times}21$ pixels window, $119{\times}119$ pixels window and $51{\times}51$ pixels window in the water area, urban land, forest land and agricultural land, respectively. And hence, DEM is generated from IKONOS stereo imagery using the selected matching sizes and land cover map on the four types.

• 치과방사선
• /
• 제28권1호
• /
• pp.17-26
• /
• 1998

The purpose of this study was to investigate whether a radiographic estimate of osseous fractal dimension is useful in the characterization of structural changes in bone. Ten specimens of bone were progressively decalcified in fresh 50 ml solutions of 0.1 N hydrochloric acid solution at cummulative timed periods of 5, 10, 20, 30, 60 and 90 minutes, and radiographed from 0 degree projection angle controlled by intraoral parelleling device. The test set of 70 radiographs was digitized and digitally filtered to reduce film -grain noise. I performed one-dimensional variance and fractal analysis of bony profiles or scan lines. Correlation analysis quantified the relationship between variance and fractal dimension. The obtained results were as follow. 1. After the first stage of decalcification variance and fractal dimension of scan line pixel intensities generally decreased with a range of 57.94 to 12.64 and 1.59 to 1.36. 2. Correlation coefficient(r) relating variances to fractal dimensions was consistantly excellent(range r=0.90 to 0.98). 3. Variance and fractal dimension were much alike in ability to discriminate, at leat on a group basis, between control and decalcified specimens.

• 한국통신학회논문지
• /
• 제24권12B호
• /
• pp.2401-2406
• /
• 1999

압축된 비디오 데이터에서 내용기반 컷 검출을 위해 다양한 특징 벡터 추출 방법이 연구되고 있다. 특징 벡터로써 화소값의 히스토그램을 이용한 방법의 경우 화소의 공간적 특성을 고려하지 않아 정확히 컷 검출을 기대하기 어렵다. 그래서 CCV(Color Coherent Vector)나 Color Correlogram등의 계산량이 복잡한 알고리즘이 많이 사용된다. 그러나 이러한 기법들은 정확한 컷 검출을 가능하게 하나 계산량이 너무 복잡하기 때문에 검출 시간이 많이 소요되는 단점이 있다. 본 논문에서는 MPEG 비디오 시퀀스에서 휘도성분의 DC값들의 공간적 상관도를 이용한 컷 검출 기법을 제안한다. 이 기법은 비교적 간단하기 때문에 처리 시간이 빠르며 또한 개선된 특징차 비교방법을 이용하여 검출율을 더 높일 수 있다.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제5권7호
• /
• pp.333-338
• /
• 2016

본 연구의 목적은 수삼 등급을 판정하는 기준 중에서 수삼의 중량을 직접 측정하지 않고 영상 분석으로 가능한 근접하게 추정하는 것이다. 이를 위해 수삼영상 취득장치를 제작하였으며 126개의 수삼샘플을 대상으로 영상을 취득하였다. 각각의 수삼샘플의 중량을 측정하고 영상분석 데이터를 이용하여 중량추정공식을 산출하는데 이용하였다. 영상분석과 파라미터 추출과정에는 C언어 기반의 Labwindows/CVI 개발 툴과 오픈소스 라이브러리 OpenCV를 이용하였다. 영상분석 과정에서 추출한 파라미터와 중량과의 상관관계를 가장 잘 표현할 수 있는 필터설정 값을 추적하여 적용하였고, 최소제곱법을 사용한 선형 회귀분석으로 선형성을 가지는 0.9162의 강한 양의 상관계수 값을 얻을 수 있었다.

• 대한원격탐사학회지
• /
• 제20권1호
• /
• pp.1-11
• /
• 2004

Based on surface energy balance climatology, surface temperatures should respond to drying conditions well before plant response. To test this hypothesis, land surface temperatures (LST) derived from MODIS data were analyzed to determine how the data were correlated with climatic water balance variables and NDVI anomalies during a growing season in Western and Central Kansas. Daily MODIS data were integrated into weekly composites so that each composite data set included the maximum temperature recorded at each pixel during each composite period. Time-integrated, or cumulative values of the LST deviation standardized with mean air temperatures had significantly high correlation coefficients with SM, AE/PE, and MD/PE, ranging from 0.65 to 0.89. The Standardized Thermal Index (STI) is proposed in this study to accomplish the objective. The STI, based on surface temperatures standardized with observed mean air temperatures, had significant temporal relationships with the hydroclimatological factors. STI classes in all the composite periods also had a strong correlation with NDVI declines during a drought episode. Results showed that, based on LST, air temperature observations, and water budget analysis, NDVI declines below normal could be predicted as early as 8 weeks in advance in this study area.

• 대한임베디드공학회논문지
• /
• 제17권6호
• /
• pp.309-317
• /
• 2022

In this paper, we propose a machine learning-based copy-move forgery detection network with dual branches. Because the rotation or scaling operation is frequently involved in copy-move forger, the conventional convolutional neural network is not effectively applied in detecting copy-move tampering. Therefore, we divide the input into rotation-invariant and scaling-invariant features based on the wavelet coefficients. Each of the features is input to different branches having the same structure, and is fused in the combination module. Each branch comprises feature extraction, correlation, and mask decoder modules. In the proposed network, VGG16 is used for the feature extraction module. To check similarity of features generated by the feature extraction module, the conventional correlation module used. Finally, the mask decoder model is applied to develop a pixel-level localization map. We perform experiments on test dataset and compare the proposed method with state-of-the-art tampering localization methods. The results demonstrate that the proposed scheme outperforms the existing approaches.

• 천문학회보
• /
• 제44권2호
• /
• pp.70.4-70.4
• /
• 2019

Image super-resolution (SR) is a technique that enhances the resolution of a low resolution image. In this study, we use three SR models (RCAN, ProSRGAN and Bicubic) for enhancing solar SDO/HMI magnetograms using deep learning. Each model generates a high resolution HMI image from a low resolution HMI image (4 by 4 binning). The pixel resolution of HMI is about 0.504 arcsec. Deep learning networks try to find the hidden equation between low resolution image and high resolution image from given input and the corresponding output image. In this study, we trained three models with HMI images in 2014 and test them with HMI images in 2015. We find that the RCAN model achieves higher quality results than the other two methods in view of both visual aspects and metrics: 31.40 peak signal-to-noise ratio(PSNR), Correlation Coefficient (0.96), Root mean square error (RMSE) is 0.004. This result is also much better than the conventional bi-cubic interpolation. We apply this model to a full-resolution SDO/HMI image and compare the generated image with the corresponding Hinode NFI magnetogram. As a result, we get a very high correlation (0.92) between the generated SR magnetogram and the Hinode one.