• Proceedings of the IEEK Conference
• /
• 2003.07e
• /
• pp.1976-1979
• /
• 2003

본 논문에서는 연속영상에서 잡음과 객체가 잘 분할되지 않는 환경 내에 있는 객체를 자동으로 분할하는 차영상 기반 알고리즘을 제안하였다. 기존의 차영상 기반의 단일 임계간을 이용한 방식에는 잡음에 크게 영향을 받고 배경과 객체가 비슷한 밝기 값을 가지는 경우 잘 추출되지 않는 많은 문제점이 있다. 본 논문에서는 이러한 문제점을 해결하고자 임계값을 설정하는 영역을 축소하여 잡음간섭의 최소화를 구성하였고 축소된 영역 내의 윤곽선정보를 이용하여 배경 밝기 값의 유사함에서 나오는 간섭을 최소화함으로써 정밀한 객체를 추출할 수 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.356-358
• /
• 2000

본 논문에서는 인간의 시각적인 감각을 멀티미디어 컨텐츠의 UMA 변환 서비스에 적용하여 영상의 다양한 디스플레이 크기의 사용자 단말기에 맞게 효율적으로 변화(해상도 축소나 Cropping) 기법을 제안하고 있다. 즉, 영상의 중요한 객체를 사각형 경계박스로 표시한 후 각각의 객체의 저자의 의도대로 사용자가 지각적으로 인식할 수 있는 최소의 공간 해상도 축소 정보를 정의하여, 영상의 변환 시 각각 객체를 사용자가 충분히 인식할 수 있는 한계치로 사용하여 효율적인 UMA 서비스를 보장하는 사용자 자원 재분배 기법을 제안한다. 또한, 본 논문에서 제안된 알고리즘을 기존의 방식과 비교하여 실험적으로 그 장단점을 비교한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1997.11a
• /
• pp.71-74
• /
• 1997

본 논문은 일반적으로 제약 없는 형식 문서 즉, 논-맨하탄(non-manhattan) 형식의 이진문서영상을 분석하는 기법으로서, 연결요소기법에 기반한 특징추출과 이를 이용한 영역분리 및 분류에 관한 새로운 방법을 제안한다. 제안한 방식은 바텀-업(bottom-up)방식으로서 먼저 처리속도의 고속화와 축소시 특징 영역보존을 위해 임계치 축소기법을 사용하고, 축소된 이진 문서영상내의 각 연결된 검은 화소의 집합을 개체화하고 개체의 특성에 따라 텍스트, 신성분, 해프톤, 도형 그리고 표 등으로 분류한다. 영역분류는 두단계로 이루어지는데, 1차분류에서는 우선, B/W 비, 면적, 외각 테두리의 높이와 너비 비, 테두리선유무 등의 특징을 이용하여 해프톤, 수평 수직선, 테두리(표 및 도형)영역을 분리한다. 이후 2차 분류에서는 문자성분의 수평결합을 통한 텍스트행 성분을 추출한다. 마지막 후처리 과정으로 표분석 알고리듬을 통하여 테두리 영역중 표와 도형을 정확히 구분하고, 또한 도형에 관련한 문서성분을 해당 도형 개체에 연결하는 작업을 수행함으로써 완벽한 영역분류를 한다. 다양한 문서영상을 이용한 시뮬레이션을 통해 제안한 알고리듬의 성능을 입증한다.

• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.959-971
• /
• 2019

CNN (Convolutional Neural Network) is one representative deep learning algorithm, which can extract high-level spatial and spectral features, and has been applied for hyperspectral image classification. However, one significant drawback behind the application of CNNs in hyperspectral images is the high dimensionality of the data, which increases the training time and processing complexity. To address this problem, several CNN based hyperspectral image classification studies have exploited PCA (Principal Component Analysis) for dimensionality reduction. One limitation to this is that the spectral information of the original image can be lost through PCA. Although it is clear that the use of PCA affects the accuracy and the CNN training time, the impact of PCA for CNN based hyperspectral image classification has been understudied. The purpose of this study is to analyze the quantitative effect of PCA in CNN for hyperspectral image classification. The hyperspectral images were first transformed through PCA and applied into the CNN model by varying the size of the reduced dimensionality. In addition, 2D-CNN and 3D-CNN frameworks were applied to analyze the sensitivity of the PCA with respect to the convolution kernel in the model. Experimental results were evaluated based on classification accuracy, learning time, variance ratio, and training process. The size of the reduced dimensionality was the most efficient when the explained variance ratio recorded 99.7%~99.8%. Since the 3D kernel had higher classification accuracy in the original-CNN than the PCA-CNN in comparison to the 2D-CNN, the results revealed that the dimensionality reduction was relatively less effective in 3D kernel.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.4
• /
• pp.271-280
• /
• 2004

This paper presents a new iterative image restoration algorithm with removal of boundary/object-oriented ringing, The proposed method is based on CGM(Conjugate Gradient Method) iterations with inter-wavelet shrinkage. The proposed method provides a fast restoration as much as CGM, while having adaptive do-noising and do-ringing by using wavelet shrinkage. In order to have effective do-noising and do-ringing simultaneously, the proposed method uses a space-dependent shrinkage rule. The improved performance of the proposed method over more traditional iterative image restoration algorithms such as LR(Lucy-Richardson) and CGM in do-noising and do-ringing is shown through numerical experiments.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.4 no.4
• /
• pp.391-397
• /
• 2003

This paper is about the digital image decimation algorithm which generates a value of decimated element by an average of a target pixel value and a value of neighbor intelligible element to adaptively reflect the merits of ZOD method and FOD method on the decimated image. First, a target pixel located at the center of sliding window is selected, then the gradient amplitudes of its right neighbor pixel and its lower neighbor pixel are calculated using first order derivative operator respectively. Secondly, each gradient amplitude is divided by the summation result of two gradient amplitudes to generate each intelligible weight. Next, a value of neighbor intelligible element is obtained by adding a value of the right neighbor pixel times its intelligible weight to a value of the lower neighbor pixel times its intelligible weight. The decimated image can be acquired by applying the process repetitively to all pixels in input image which generates the value of decimated element by calculating the average of the target pixel value and the value of neighbor intelligible element.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.10
• /
• pp.2091-2101
• /
• 1997

Conventional method for fractal image decoding requires high-degree computational complexity in decoding propocess, because of iterated contractive transformations applied to whole range blocks. In this paper, we propose a fast decoding algorithm of fractal image using data depence in order to reduce computational complexity for iterated contractive transformations. Range of reconstruction image is divided into a region referenced with domain, called referenced range, and a region without reference to domain, called unreferenced range. The referenced range is converged with iterated contractive transformations, and the unreferenced range can be decoded by convergence of the referenced range. Thus the unreferenced range is called data dependence region. We show that the data dependence region can be deconded by one transformation when the referenced range is converged. Consequently, the proposed method reduces computational complexity in decoding process by executing iterated contractive transformations for the referenced range only.

• Journal of Broadcast Engineering
• /
• v.10 no.3
• /
• pp.413-421
• /
• 2005

As the population of HDTV is growing, among many useful features of modern set top boxes (STBs) or digital video recorders (DVRs), video browsing, visual bookmark, and picture-in-picture capabilities are very frequently required. These features typically employ reduced-size versions of video frames, or thumbnail images. Most thumbnail generation approaches generate DC images directly from a compressed video stream. A discrete cosine transform (DCT) coefficient for which the frequency is zero in both dimensions in a compressed block is called a DC coefficient and is simply used to construct a DC image. If a block has been encoded with field DCT, a few AC coefficients are needed to generate the DC image in addition to a DC coefficient. However, the bit length of a codeword coded with variable length coding (VLC) cannot be determined until the previous VLC codeword has been decoded, thus it is required that all codewords should be fully decoded regardless of their necessary for DC image generation. In this paper, we propose a method especially for fast DC image generation from an I-frame using multiple-symbol lookup table (mLUT). The experimental results show that the method using the mLUT improves the performance greatly by reducing LUT count by 50$\%$.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.8
• /
• pp.1714-1719
• /
• 2011

High quality image processing schemes are used more widely than ever according to the development of various visual media. We need a compressed form of image for sending more capacity and a controlling strategy of images for small display devices. In this paper, we propose an image upsampling and downsamplig scheme using DCT coefficients for those purposes. Our scheme is designed to control the size of picture based on the target display media by reducing the data in DCT domain while not increasing the computational burdens. With the power of controlling the resolution in DCT domain, the proposed method shows higher PSNR than other competing methods in experiment.

• Journal of Digital Contents Society
• /
• v.4 no.2
• /
• pp.211-221
• /
• 2003

The objective of this paper is performance analysis of the digital document image decimation algorithm which generates a value of decimated element by an average of a target pixel value and a value of neighbor intelligible element to adaptively reflect the merits of ZOD method and FOD method on the decimated image. First, a target pixel located at the center of sliding window is selected, then the gradient amplitudes of its right neighbor pixel and its lower neighbor pixel are calculated using first order derivative operator respectively. Secondly, each gradient amplitude is divided by the summation result of two gradient amplitudes to generate each local intelligible weight. Next, a value of neighbor intelligible element is obtained by adding a value of the right neighbor pixel times its local intelligible weight to a value of the lower neighbor pixel times its intelligible weight. The decimated image can be acquired by applying the process repetitively to all pixels in input image which generates the value of decimated element by calculating the average of the target pixel value and the value of neighbor intelligible element. In this paper, the performance comparison of proposed method and conventional methods in terms of subjective performance and hardware complexity is analyzed and the preferable approach for developing the decimation algorithm of the digital document image on the basis of this analysis result has been reviewed.