Fractals을 이용한 Very Low Bit Rate 동화상 영상코더에 관한 연구 ( Very Low Bit Rate Video Coder Using the Fractals )

장종환,정재길,박두영 대한전자공학회 1995 대한전자공학회 학술대회 Vol.1995 No.1

입술형태 특성을 이용한 음성코딩

장종환,Jang, Jong-Hwan 배재대학교 공학연구소 1997 공학논문집 Vol.1 No.1

음성을 전송하는데에 있어서 여러 가지 제약이 있는 경우에 더 좋은 방법으로 말하는 사람의 입을 관찰하여 입모양이 나타내는 특징 값들을 이용해 음성을 알아내고 이미 저장된 Database에서 특징 값에 해당하는 코드를 상대방에 전송하는 것이다. 실제 음성을 전송하지 않기 때문에 신호에 대한 잡음이나 보안문제를 해결할 수 있다. To transmit the degraded voice signal within various environment surrounding acoustic noises, we extract lip i the face and then compare lip edge features with prestoring DB having features such as mouth height, width, area, and rate. It provides high security and is not affected by acoustic noise because it is not necessary to transmit the actual utterance.

텍스처 세그멘티이션 영상코딩 기술의 성능분석

장종환,정재길,양우석 대한전자공학회 1992 대한전자공학회 학술대회 Vol.1992 No.1

Effectiveness of Transfer Learning for Deep Learning-Based Electrocardiogram Analysis

장종환,김태영,윤덕용 대한의료정보학회 2021 Healthcare Informatics Research Vol.27 No.1

Objectives: Many deep learning-based predictive models evaluate the waveforms of electrocardiograms (ECGs). Becausedeep learning-based models are data-driven, large and labeled biosignal datasets are required. Most individual researchersfind it difficult to collect adequate training data. We suggest that transfer learning can be used to solve this problem and increasethe effectiveness of biosignal analysis. Methods: We applied the weights of a pretrained model to another model thatperformed a different task (i.e., transfer learning). We used 2,648,100 unlabeled 8.2-second-long samples of ECG II data topretrain a convolutional autoencoder (CAE) and employed the CAE to classify 12 ECG rhythms within a dataset, which had10,646 10-second-long 12-lead ECGs with 11 rhythm labels. We split the datasets into training and test datasets in an 8:2 ratio. To confirm that transfer learning was effective, we evaluated the performance of the classifier after the proposed transferlearning, random initialization, and two-dimensional transfer learning as the size of the training dataset was reduced. Allexperiments were repeated 10 times using a bootstrapping method. The CAE performance was evaluated by calculating themean squared errors (MSEs) and that of the ECG rhythm classifier by deriving F1-scores. Results: The MSE of the CAE was626.583. The mean F1-scores of the classifiers after bootstrapping of 100%, 50%, and 25% of the training dataset were 0.857,0.843, and 0.835, respectively, when the proposed transfer learning was applied and 0.843, 0.831, and 0.543, respectively, afterrandom initialization was applied. Conclusions: Transfer learning effectively overcomes the data shortages that can compromiseECG domain analysis by deep learning.

A High Image Compression for Computer Storage and Communication

장종환,Jang, Jong-Whan 배재대학교 자연과학연구소 1991 自然科學論文集 Vol.4 No.-

Human Visual System(HVS)의 특성과 image의 textural regions의 roughness을 이용하여 image segmentation을 행하여 high compression에서도 고화질을 나타내는 새로운 image coder를 이 논문에서 논한다. 제안된 image coder는 constant segments를 가진 segmentation-based image coding technique의 문제들을 다음과 같은 방법론을 제안함으로써 해결하였다. Image를 HVS으로 보았을 때 degree of roughness에 관하여 textually homogeneous regions으로 segmentation하였다. Fractal dimension을 roughness of textural regions을 측정하기 위하여 사용하였다. Segmentation은 fractal dimension을 thresholding하여 textural regions이 three texture classes로 분류하였다(perceived constant intensity, smooth texture, and rough texture). High compression을 가지는 고질화의 image coder는 각각의 segment boundary와 각각의 texture class에 효율적인 coding technique를 적용 함으로 얻었다. A new texture segmentation-based image coding technique which performs segmentation based on roughness of textural regions and properties of the human visual system (HVS) is presented. This method solves the problems of a segmentation-based image coding technique with constant segments by proposing a methodology for segmenting an image texturally homogeneous regions with respect to the degree of roughness as perceived by the HVS. The fractal dimension is used to measure the roughness of the textural regions. The segmentation is accomplished by thresholding the fractal dimension so that textural regions are classified into three texture classes; perceived constant intensity, smooth texture, and rough texture. An image coding system with high compression and good image quality is achieved by developing an efficient coding technique for each segment boundary and each texture class. For the boundaries, a binary image representing all the boundaries is created. For regions belonging to perceived constant intensity, only the mean intensity values need to be transmitted. The smooth and rough texture regions are modeled first using polynomial functions, so only the coefficients characterizing the polynomial functions need to be transmitted. The bounda-ries, the means and the polynomial functions are then each encoded using an errorless coding scheme. Good quality reconstructed images are obtained with about 0.08 to 0.3 bit per pixel for three different types of imagery ; a head and shoulder image with little texture variation, a complex image with many edges, and a natural outdoor image with highly textured areas.

강건한 이미지 초해상도를 위한 적응형 가중치 맵을 사용한 메타 학습

장종환,최장훈 대한전자공학회 2024 전자공학회논문지 Vol.61 No.3

이미지 초해상도(SR)는 저해상도 이미지를 고해상도 이미지로 변환하는 컴퓨터 비전 기술이다. 딥러닝의 등장으로 새롭고 효과적인 SR 방법들이 많이 제안되었지만, 현재 대부분의 SR 방법들은 저해상도 이미지의 저하 과정이 bicubic downsampling이라고 가정하므로, 여러 내/외부 요인으로 다양한 저하 유형을 가지는 실제 이미지에는 잘 적용되지 않는 문제가 있다. 본 논문에서는 SR 네트워크 구조를 변경하지 않고, 입력 이미지의 내부 정보를 활용해 단일 이미지 초해상도 성능을 개선하는 방법을 제안한다. 이미지 내부 정보를 활용하기 위해 입력 이미지에 특화된 적응형 픽셀 단위 가중치 맵(PAW)을 구축하는 메타 러너 네트워크를 사용하는 메타 학습 방법을 사용한다. 이 접근법은 bicubic downsampling뿐만 아니라, blind downsampling에도 빠르게 적응 가능하며, 네트워크가 이미지의 복잡한 부분에 더 집중하도록 유도함으로써 성능을 향상시킨다. 다양한 벤치마크 SR 데이터 세트에서의 실험 결과, 제안 방법이 기존 SR 네트워크의 구조를 유지하면서 성능을 향상시키는 것을 확인했다. Image super-resolution(SR) is a computer vision task that converts low-resolution images into high-resolution images. With the advent of deep learning, many new and effective methods have been proposed. However, most SR methods are conducted under the assumption that the low-resolution degradation process is bicubic downsampling, so there is a problem that they are not well applied to real images with complex degradation types. In this work, we propose a method to improve the performance of Single Image Super-Resolution that is robust against various degradation types, without changing the architecture of conventional SR network, by utilizing the internal information. For utilizing the internal information, we adopt a meta-learning algorithms using a meta-learner network, which constructs an pixel-wise adaptive weight map(PAW) tailored to the given input image. This approach can be quickly applied to bicubic downsampling kernels as well as blind downsampling kernels and improves performance by encouraging the network to focus more on complex parts of the image. Experiments on various benchmark SR datasets show that our proposed method improves the performance while maintaining the structure of the existing SR network.

복수객체의 윤곽추출 알고리즘 성능분석

장종환 배재대학교 공학연구소 2011 공학논문집 Vol.13 No.1

The most important operation in Snake algorithms of extracting boundary of multiple objects is to find the segment intersection and to split and connect segments appropriately. In most of previous algorithms, two 1-D equations are solved to find segment intersection and the nearest distance method is used for splitting and connecting snake points. This method is time-consuming. Connecting snake points using the distance between snake points creates the wrong connection. To solve these problems, new splitting and connecting operation using vectors of snake segments is proposed.

Gunnar Myrdal의 福祉國家觀에 對한 一考察

장종환 원광대학교 사회복지학과 1987 圓光社會福祉學 Vol.2 No.-

직접 주파수 합성에 관한 연구 ( A Study on the Direct Frequency Synthesis )

장종환,한영열 한국통신학회 1984 한국통신학회 학술대회논문집 Vol.3 No.1

MPEG 압축도매인상에서 HVS을 이용한 장면변환 추출 알고리즘 ( An algorithm of detecting scene change using HVS in the MPEG compressed domain )

장종환,이해문 대한전자공학회 1998 대한전자공학회 학술대회 Vol.1998 No.1