• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2012.11a
• /
• pp.25-28
• /
• 2012

본 논문에서는 키넥트(Kinect)에서 획득한 깊이 영상의 품질을 향상시키는 기법을 제안한다. 키넥트는 마이크로소프트사에서 출시한 카메라로 깊이 영상과 컬러 영상을 획득 할 수 있다. 하지만 적외선 패턴을 이용한 깊이 영상의 획득 방법의 한계로 인해 객체의 경계 주변으로 홀 및 잡음이 생긴다. 따라서 정확한 깊이 영상을 얻기 위해서는 깊이 영상의 품질 향상이 필수적이다. 일반적으로 깊이 영상의 홀을 채울 때, 인페인팅(inpainting) 또는 결합형 양방향 필터(joint bilateral filter) 등의 기법을 사용한다. 그러나 이러한 기법들의 경우 한 장의 영상만을 이용하기 때문에 객체 경계 주변의 흔들림 현상을 보정할 수 없다. 제안하는 기법에서는 먼저 수행속도가 빠른 가우시안 필터를 이용하여 경계 주변의 홀을 채운다. 이전 프레임의 컬러 영상을 그레이 영상으로 변환한 다음에 그레이 영상과 깊이 영상의 값의 변화를 분석하여 흔들림 화소를 찾아 이전 깊이 영상들 중 최대 화소 값으로 변환함으로써 깊이 영상의 경계 흔들림 현상을 줄일 수 있다. 실험을 통해 제안하는 기법이 기존의 방법들 보다 우수하다는 것을 확인하였다.

• The Journal of the Korea Contents Association
• /
• v.7 no.9
• /
• pp.72-80
• /
• 2007

Image magnification is the estimation of a few pixel in images with high quality from a pixel of an image with low resolution and there have been studied many techniques to make images with high quality. In this paper, we propose an image interpolation method using median filter and spatial information. The proposed method makes an interpolating pixel using an average value of a median filtered value and an average value of two pixels correlated with an interpolating pixel tightly. Also we make the magnified image with improved quality to add the directional information of surrounding pixels and the characteristic of ones using average value and max value of spatial variation. We evaluate the performance using PSNR in the quality of enlarged image comparing the proposed method with existing methods. The results show the proposed method improves PSNR than the existing methods and make images preserving the characteristic of original imges.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.2
• /
• pp.207-213
• /
• 2014

In this paper, we propose a modified median filter for impulse noise reduction. The proposed method based on multi-step finds noisy pixels from the corrupted image and applies filtering on the noisy pixels. Neighbor pixels for filtering are filtered by linear filter which adjusts filtering direction according to an edge. Thus, our proposed method not only preserves edge, but also reduces noise in uniform region. Experimental results show that our proposed method has better quality than those by existing modified median filtering method.

• Journal of Korea Multimedia Society
• /
• v.6 no.2
• /
• pp.224-229
• /
• 2003

In this paper, an effective skeleton method is proposed in order to obtain an enhanced digital image of skeleton line. The edge-detection method is applied in the preprocessing stage and after that, the modified Parallel method is applied to obtain the improved image of skeleton line. The existing parallel methods are Zhang, Lu and Wang, and Paul methods. Firstly, a parallel process method Is applied, and the proposed method is applied that the original is compared with the four neighbor pixels and four corner pixels of mask. In conclusion, the proposed method shows an improved connectivity and quality of skeleton line.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2011.11a
• /
• pp.298-301
• /
• 2011

Edge detection is an base research task in the field of image processing. Edge detection can be regarded as a technique for locating pixels of abrupt gray-level change. So with Gradient method, it can be computed easily. But it can't satisfy human naked eye. so in this paper, new algorithm based on the NLM(Non Local Means) is proposed for good performance for human naked eye.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2002.11b
• /
• pp.234-237
• /
• 2002

디지털 워터마크는 멀티미디어 콘텐츠에 대한 저작권 보호와 인증의 목적으로 최근에 제안되었다. 워터마킹 기술은 공간영역이나 주파수 영역에서 워터마크를 삽입하는 알고리즘으로 구성된다. 본 논문에서는 기존의 공간영역의 워터마킹 방법을 보안, 개선하였고, 주변화소의 밝기를 고려하여 워터마크의 정보를 삽입하는 방법을 이용하였다. 실험은 영상의 화질저하를 추정하기 위해 PSNR을 계산하고, 제안하는 알고리즘으로 삽입된 영상의 화질이 개선됨을 보이고 외부의 공격에도 워터마크가 추출됨을 보인다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.12C
• /
• pp.738-743
• /
• 2011

For the high quality 3-D broadcasting, depth maps are important data. Although commercially available depth cameras capture high-accuracy depth maps in real time, their resolutions are much smaller than those of the corresponding color images due to technical limitations. In this paper, we propose the depth map up-sampling method using a high-resolution color image and a low-resolution depth map. We define a random walk probability model in an operation unit which has nearest seed pixels. The proposed method is appropriate to match boundaries between the color image and the depth map. Experimental results show that our method enhances the depth map resolution successfully.

• Proceedings of the Korean Information Science Society Conference
• /
• 2010.06c
• /
• pp.226-229
• /
• 2010

본 연구에서는 다양한 density profile 특징추출에 기반한 염색체 자동분류방법들의 성능을 비교분석하였다. density profile은 염색체의 밴드패턴을 가장 잘 표현한 특징으로 염색체의 중심축을 구성하는 화소들의 밝기 값을 추출하는 방법이다. 염색체의 밴드패턴은 염색체의 끝단까지를 잘 표현해주어야만 정확한 염색체번호를 확인할 수 있다. 따라서 염색체의 중심축을 추출하여 염색체 끝단까지 확장 처리한 방법에 대한 성능을 확인하였다. 염색체 중심축에 위치한 화소만을 이용한 프로파일은 잡음에 민감할 수 있으므로 이를 해결하기 위하여 염색체의 중심축에 대한 화소 값 대신 주변 밝기 값들에 대한 평균을 이용한 국소평균방법과 중심축의 수직라인 상에 존재하는 화소 값들에 대한 평균을 구한 수직평균방법을 비교하였다. 분류알고리즘은 k-NN을 사용하였고, 실험데이터는 (주)Gendix 로부터 제공받은 임상적으로 정상인 100명(남자 50명, 여자 50명)으로부터 추출한 4600개의 염색체 영상을 훈련데이터와 테스트데이터로 각각 50%씩 랜덤하게 분리하여 실험하였다. 실험결과 중심축을 확장하고 수직평균에 대한 프로파일을 특징으로 추출하여 분류한 경우가 가장 좋은 성능을 보였다.

• The Journal of the Korea Contents Association
• /
• v.7 no.12
• /
• pp.346-354
• /
• 2007

Interpolation methods to get the magnified image from an image with low resolution use known pixels to make an interpolated pixel. This interpolation process usually generates blurred edges and blocking effect in the result image. To improve these defects, conventional methods multiply proper weights reflecting neighborhood pixels and add the values during interpolating process. The proposed method changes input pixels in consideration of information of neighborhood pixels, gets interpolated pixels by using these values and improves the quality of interpolated image. Firstly, we compute difference values of the diagonal directions of a pixel and classify flat regions and complex regions according to these values. If the regions is complex ones, the proposed method changes an original pixel into a new value using a input pixel and a median value of it's neighbor pixels. Therefore, the proposed method applies bilinear method to the original pixels in flat regions and the changed ones in complex regions and produces the interpolated images. We evaluate the performance of the proposed method with existing methods by using PSNR and the quality of enlarged image. The results show that the proposed method improves PSNR in comparing with conventional methods and that is superior to the existing methods in terms of the quality of the interpolated image.

• Journal of Biomedical Engineering Research
• /
• v.23 no.3
• /
• pp.253-261
• /
• 2002

A spatially adaptive falter was formulated to imrove PET image qualify and the Performance of the filter was evaluated using simulation and phantom and human PET studies. In the proposed filter. if a pixel was identified as the edge Pixel, the Pixel value was Preserved. Otherwise a Pixel was replaced by the mean of the pixel values weighted by 2:7: 2. A Pixel was identified as the edge Pixel. if it satisfies the following conditions : the number of ADs (absolute difference between center and neighborhood pixels) which is smaller than THl (($pix_max{\times}0.1/log_2(NPM)$, NPM : mean of 6 neighborhood pixels excluding minimum and maximum) is 8-k and the number of ADs which is lager than TH2 ($NPM{\times}0.1$) is k. where k : 2, 3, …, 6. The results of this study demonstrate the superior performance of the Proposed titter compared to Gaussian fitter, weight median filter and subset averaged median filter. The proposed tittering method is simple but effective in increasing uniformity and contrast with minimal degradation of spatial resolution of PET images and thus. is expected to Provide improved diagnositc quality PET images .