• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2007.10a
• /
• pp.413-417
• /
• 2007

2006년 11월 이후 신 차량 번호판 등장 후, 신 차량 번호판 차량이 꾸준히 증가하고 있다. 이에 따라 속도위반, 신호위반 단속, 무인 주차 관리 시스템, 범죄 및 도주 차량 검거, 고속도로 톨게이트에서 통행료 지불로 인한 교통 체증현상을 해소하기 위한 자동 요금 징수와 같은 다양한 경우에서 신 자동차 번호판의 특징에 맞는 인식 시스템이 요구되고 있다. 따라서 본 논문에서는 이러한 문제를 해결하기 위해 지능형 신 자동차 번호판 인식 방법을 제안한다. 무인 카메라에서 획득된 신 차량 영상을 그레이 레벨로 변환한 후에 블록 이진화한다. 블록 이진화된 차량 영상을 대상으로 차량의 형태학적 특징을 적용하여 잡음을 제거한 후, 번호판 영역을 추출한다. 추출된 번호판 영역에 대해 Grassfire 알고리즘을 적용하여 개별 코드를 추출한다. 차량 번호판을 인식하기 위하여 추출된 개별 코드를 퍼지 ART 알고리즘을 적용하여 학습 및 인식한다. 제안된 차량 번호판 추출 및 인식 방법의 성능을 평가하기 위해 100장의 차량 영상을 대상으로 실험한 결과, 제안된 차량 번호판 추출 및 인식 방법이 실험을 통해서 효율적인 것을 확인하였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.6
• /
• pp.1281-1285
• /
• 2013

In this paper, we propose a method to extract the area of appendix from ultrasonographic image via computational vision. A series of image processing techniques such as Ends-in search stretching for enhancing the brightness contrast, block binarization, grassfire algorithm for extracting lower part of fascia, and fuzzy sigma binarization method to finalize the appendix area are used to achieve our goal. The strength of this paper is using fuzzy sigma binarization instead of other binarization technique to handle the sensitivity of extracting the target area from regio hypogastrica. The experiment verifies the efficacy of the proposed method successfully.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.662-663
• /
• 2020

최근 MPEG 에서는 인공신경망 모델을 다양한 딥러닝 프레임워크에서 상호운용 가능한 포맷으로 압축 표현할 수 있는 NNR(Compression of Neural Network for Multimedia Content Description and Analysis) 표준화를 진행하고 있다. 본 논문에서는 MPEG-NNR 에서 CNN 모델을 압축하기 위한 지역 비선형 양자화(Local Non-linear Quantization: LNQ) 기법을 제시한다. 제안하는 LNQ 는 균일 양자화된 CNN 모델의 각 계층의 가중치 행렬 블록 단위로 추가적인 비선형 양자화를 적용한다. 또한, 제안된 LNQ 는 가지치기(pruning)된 모델의 경우 블록내의 영(zero) 값의 가중치들은 그대로 전송하고 영이 아닌 가중치만을 이진 군집화를 적용한다. 제안 기법은 음성 분류를 위한 CNN 모델(DCASE Task)의 압축 실험에서 기존 균일 양자화를 대비 동일한 분류 성능에서 약 1.78 배 압축 성능 향상이 있음을 확인하였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.5
• /
• pp.540-546
• /
• 2019

This study suggests a binarization method that minimizes information loss for night illumination images. The object of the night illumination image is an image which is not focused due to the influence of illumination and is not identifiable. Also, the image has a brightness area in only a part of the brightness histogram. So the existing simple binarization method is hard to get good results. The proposed binarization method uses image segmentation method and image merging method. In the stepwise divided blocks, we divide into two regions using the triangular type of fuzzy logic. The value 0 of the membership degree is binarized at the present step, and the value of the membership degree 1 is binarized after the next step. Experimental results show that night illumination images with minimal loss of information can be obtained in a dark area brightness range.

• The KIPS Transactions:PartB
• /
• v.18B no.5
• /
• pp.265-270
• /
• 2011

In this paper, we propose a method for intensity correction using binarization-based region segmentation in 3D stereoscopic images. In the proposed method, 3D stereoscopic right image is segmented using binarizarion. Small regions in the segmented image are eliminated. For each region in right image, a corresponding region in left image is decided through region matching using correlation coefficient. When region-based matching, in order to prevent overlap between regions, we remove a portion of the area closed to the region boundary using morphological filter. The intensity correction in left and right image can be performed through histogram specification between the corresponding regions. Simulation results show the proposed method has the smallest matching error than the conventional method when we generate the right image from the left image using block based motion compensation.

• Journal of the Korea Society of Computer and Information
• /
• v.13 no.6
• /
• pp.273-278
• /
• 2008

Cars attaching new license plates are increasing after introducing the new format of car license plate in Korea. Therefore, a car new license plate recognition system is required for various fields using automatic recognition of car license plates, automatic parking management systems and arrest of criminal or missing vehicles. In this paper, we proposed an intelligent new car license plate recognition method for the various fields. The proposed method is as follows. First of all, an acquired color image from a surveillance camera is converted to a gray level image and binarized by block binarization method. Second, noises of the binarized image removed by morphological characteristics of cars and then license plate area is extracted. Third, individual characters are extracted from the extracted license plate area using Grassfire algorithm. lastly, the extracted characters are learned and recognized by a fuzzy ART algorithm for final car license plate recognition. In the experiment using 100 car images, we could see that the proposed method is efficient.

• Journal of Korea Multimedia Society
• /
• v.6 no.3
• /
• pp.429-435
• /
• 2003

In this paper, we propose improved binary tree vector quantization with consideration of spatial sensitivity which is one of the human visual properties. We combine weights in consideration with the responsibility of human visual system according to changes of three primary color in blocks of images with the process of splitting nodes using eigenvector in binary tree vector quantization. Also we propose the novel quality measure of the quantization images that applies MTF(modulation transfer function) to luminance value of quantization error of color image. The test results show that the proposed method generates the quantized images with fine color and performs better than the conventional method in terms of clustering the similar regions. Also the proposed method can get less quantized level images and can reduce the resource occupied by the quantized image.

• Journal of Korea Multimedia Society
• /
• v.8 no.10
• /
• pp.1347-1359
• /
• 2005

We propose an efficient block classification of the document images using the second-order statistical texture features computed from spatial gray level dependence matrix (SGLDM). We studied on the techniques that will improve the block speed of the segmentation and feature extraction speed and the accuracy of the detailed classification. In order to speedup the block segmentation, we binarize the gray level image and then segmented by applying smoothing method instead of using texture features of gray level images. We extracted seven texture features from the SGLDM of the gray image blocks and we applied these normalized features to the BP (backpropagation) neural network, and classified the segmented blocks into the six detailed block categories of small font, medium font, large font, graphic, table, and photo blocks. Unlike the conventional texture classification of the gray level image in aerial terrain photos, we improve the classification speed by a single application of the texture discrimination mask, the size of which Is the same as that of each block already segmented in obtaining the SGLDM.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.3
• /
• pp.179-185
• /
• 2019

This paper proposes a block-type classification based image binarization for the implementation of the low-power feature extraction algorithm. The proposed method can be implemented with threshold value re-use technique approach when the image divided into $64{\times}64$ macro blocks size and calculating the threshold value for each block type only once. The algorithm is validated based on quantitative results that only a threshold value change rate of up to 9% occurs within the same image/block type. Existing algorithms should compute the threshold value for 64 blocks when the macro block is divided by $64{\times}64$ on the basis of $512{\times}512$ images, but all suggestions can be made only once for best cases where the same block type is printed, and for the remaining 63 blocks, the adaptive threshold calculation can be reduced by only performing a block type classification process. The threshold calculation operation is performed five times when all block types occur, and only the block type separation process can be performed for the remaining 59 blocks, so 93% adaptive threshold calculation operation can be reduced.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2011.04a
• /
• pp.64-67
• /
• 2011

최근 개인 프라이버시 보호에 대한 중요성이 제기되면서, 보안에 대한 관심이 증가하게 되었다. 본 논문에서는 한국 산업규격 KS 표준 블록 암호 알고리즘인 ARIA의 핵심논리를 유지하면서, 동종 경쟁 알고리즘(AES, Camellia 등)과의 차별성으로 강조되어온 16 ${\times}16$ 이진 행렬을 이용한 확산 계층을$ (4{\times}4){\times}4$의 이진 행렬 형태로 수정한 개선 ARIA를 구현하였다. 개선 설계된 ARIA를 검증하기 위해, 파일 암.복호화 시스템을 적용하였고, 보안 영상 시스템을 개발하였다. 기존의 ARIA의 장점을 유지하기 때문에, 초경량 환경이나 많은 데이터를 초고속으로 처리에 필요한 응용에 더 효과적으로 적용될 수 있다.