• 한국의학물리학회:학술대회논문집
• /
• 한국의학물리학회 1999년도 Japanese Journal of Medical Physics
• /
• pp.162-163
• /
• 1999

• 대한전자공학회논문지SP
• /
• 제41권2호
• /
• pp.41-50
• /
• 2004

MPEG-4 표준에서는 객체 단위의 부호화를 수행하기 위해 자연영상으로부터 비디오 객체를 분리하는 영상분할(segmentation) 기술이 필요하다. 영상분할 방법은 크게 자동 영상분할(automatic segmentation)과 반자동 영상분할(semi-automatic segmentation)의 두 부류로 나눌 수 있다. 지금까지 개발된 대부분의 자동 영상분할 방법은 비디오 객체의 명확한 수학적인 모델을 제시하기 곤란하며 한 화면에서 개별 객체를 추출하기 어렵기 때문에 그 성능에 한계가 있다. 본 논문에서는 이러한 문제점을 극복하기 위해 active contour 알고리즘을 이용한 반자동 영상분할 알고리즘을 제안한다. 초기 곡선으로부터 변화 가능한 모든 곡선의 집합을 모양공간으로 정의하고 그 공간을 선형공간이라고 가정하면, 모양공간(shape space)은 모양 행렬에 의해 행(column) 공간과 남은 빈(left null) 공간으로 나뉘어진다. 본 논문에서 제안하는 알고리즘은 행공간의 모양공간 벡터를 이용하여 초기 곡선으로부터 영상의 특징점까지의 변화를 기술하고 동적 그래프 검색 알고리즘을 이용하여 객체의 세밀한 부분을 묘사한다. 모양 행렬과 객체의 윤곽을 추정하기 위한 SUSAN 연산자의 사용으로 제안한 알고리즘은 저수준 영상처리로부터 생성되는 불필요한 특징점을 무시할 수 있다. 또한, 모양 행렬의 사용으로 생긴 제약은 동적 그래프 검색 알고리즘으로 보상한다.

• 한국통신학회논문지
• /
• 제39C권6호
• /
• pp.490-496
• /
• 2014

무릎 관절 연골은 두께가 얇아 대부분 무릎 질환의 원인이 되고 있다. 그러므로 무릎 자기공명영상에서 관절 연골 분할은 무릎 질환의 정확한 진단을 위한 필수조건이다. 특히 수동이 아닌 전자동 방식으로 무릎 관절 연골을 분할하여야만 효과적인 무릎 질환 진단을 할 수 있다. 본 논문에서는 뇌 자기공명영상에서 대표적으로 사용되는 레벨 셋 기반의 영상 분할 기법을 분석하여 무릎 자기공명영상에 적용 시 문제점을 파악하고 이를 해결함으로써, 무릎 자기공명영상에 레벨 셋 기반 영상분할 방식을 적용하였다. 이는 본 논문에서 제안하는 분할기법을 사용할 경우 무릎 관절 연골 분할에 대한 모든 과정이 전자동화 되어 기존 반자동화 방식보다 빠른 처리가 가능하며, 3차원 형상화를 통해 보다 정확한 진단에 도움을 줄 수 있다. 또한 우리는 제안하고 있는 분할기법이 기존 대표적인 무릎 관절 분할보다 더 높은 정확도를 갖는 것을 실험을 통해 확인할 수 있었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권1호
• /
• pp.105-125
• /
• 2024

Gliomas are the most common malignant brain tumor and cause the most deaths. Manual brain tumor segmentation is expensive, time-consuming, error-prone, and dependent on the radiologist's expertise and experience. Manual brain tumor segmentation outcomes by different radiologists for the same patient may differ. Thus, more robust, and dependable methods are needed. Medical imaging researchers produced numerous semi-automatic and fully automatic brain tumor segmentation algorithms using ML pipelines and accurate (handcrafted feature-based, etc.) or data-driven strategies. Current methods use CNN or handmade features such symmetry analysis, alignment-based features analysis, or textural qualities. CNN approaches provide unsupervised features, while manual features model domain knowledge. Cascaded algorithms may outperform feature-based or data-driven like CNN methods. A revolutionary cascaded strategy is presented that intelligently supplies CNN with past information from handmade feature-based ML algorithms. Each patient receives manual ground truth and four MRI modalities (T1, T1c, T2, and FLAIR). Handcrafted characteristics and deep learning are used to segment brain tumors in a Global Convolutional Neural Network (GCNN). The proposed GCNN architecture with two parallel CNNs, CSPathways CNN (CSPCNN) and MRI Pathways CNN (MRIPCNN), segmented BraTS brain tumors with high accuracy. The proposed model achieved a Dice score of 87% higher than the state of the art. This research could improve brain tumor segmentation, helping clinicians diagnose and treat patients.

• 대한전자공학회논문지SP
• /
• 제40권5호
• /
• pp.303-311
• /
• 2003

MPEG-4 표준에서는 객체 단위의 부호화를 수행하기 위해 우선 자연영상으로부터 비디오 객체론 분리하는 영상분할(Segmentation) 기술이 필요하다. 영상분할 방법은 크게 자동 영상분할(Automatic Segment값ion)과 반자동 영상분할(Semi-automatic Segmentation)의 두 부류로 나눌 수 있다. 대부분의 자동 영상분할 방법은 비디오 객체의 명확한 모델을 수학적으로 제시하기 어려우므로 한 화면에서 개별 객체를 추출하기 어렵기 때문에 그 성능에 한계가 있다. 본 논문에서는 이러한 문제점을 극복하기 위해 기하학적인 Active Contour를 이용한 반자동 영상분할 알고리즘을 제안한다. 매개변수 방식의 Active Contour와 달리, 기하학적인 Active Contour는 곡선의 변화론 Level Set 방법을 이용하여 기술하기 때문에 초기 곡선의 모양을 객체의 모양과 무관하게 그릴 수 있다. 평탄화된 영상으로부터 경계함수를 생성하기 위해 이진화된 3차원 확산 모델을 사용하여 LUV 벡터 공간에서 비등방형 확산을 수행한다. 본 논문에서는 흐름 벡터장(Advection Vector Field)에서 곡선을 수축하고, 움직임 정보를 이용하여 곡선 확장하는 방법을 이용하여 동영상에서 객체를 분리하는 방법을 제안한다.

• Smart Structures and Systems
• /
• 제31권4호
• /
• pp.365-381
• /
• 2023

The existing vision-based techniques for inspection and condition assessment of civil infrastructure are mostly manual and consequently time-consuming, expensive, subjective, and risky. As a viable alternative, researchers in the past resorted to deep learning-based autonomous damage detection algorithms for expedited post-disaster reconnaissance of structures. Although a number of automatic damage detection algorithms have been proposed, the scarcity of labeled training data remains a major concern. To address this issue, this study proposed a semi-supervised learning (SSL) framework based on consistency regularization and cross-supervision. Image data from post-earthquake reconnaissance, that contains cracks, spalling, and exposed rebars are used to evaluate the proposed solution. Experiments are carried out under different data partition protocols, and it is shown that the proposed SSL method can make use of unlabeled images to enhance the segmentation performance when limited amount of ground truth labels are provided. This study also proposes DeepLab-AASPP and modified versions of U-Net++ based on channel-wise attention mechanism to better segment the components and damage areas from images of reinforced concrete buildings. The channel-wise attention mechanism can effectively improve the performance of the network by dynamically scaling the feature maps so that the networks can focus on more informative feature maps in the concatenation layer. The proposed DeepLab-AASPP achieves the best performance on component segmentation and damage state segmentation tasks with mIoU scores of 0.9850 and 0.7032, respectively. For crack, spalling, and rebar segmentation tasks, modified U-Net++ obtains the best performance with Igou scores (excluding the background pixels) of 0.5449, 0.9375, and 0.5018, respectively. The proposed architectures win the second place in IC-SHM2021 competition in all five tasks of Project 2.

• International Journal of Contents
• /
• 제3권4호
• /
• pp.15-21
• /
• 2007

Image morphing deals with the metamorphosis of one image into another. The field morphing depends on the manual work for most of the process, where a user has to designate the control lines. It takes time and requires skills to have fine quality results. It is an object of this paper to propose a method capable of realizing the semi-automation of field morphing using adaptive vertex correspondence based on image segmentation. The adaptive vertex correspondence process efficiently generates a pair of control lines by adaptively selecting reference partial contours based on the number of vertices that are included in the partial contour of the source morphing object and in the partial contour of the destination morphing object, in the pair of the partial contour designated by external control points through user input. The proposed method generates visually fluid morphs and warps with an easy-to-use interface. According to the proposed method, a user can shorten the time to set control lines and even an unskilled user can obtain natural morphing results as he or she designates a small number of external control points.

• 한국멀티미디어학회논문지
• /
• 제16권9호
• /
• pp.1018-1030
• /
• 2013

The corpus callosum is the largest connective structure in the brain, and its shape and size are correlated to sex, age, brain growth and degeneration, handedness, musical ability, and neurological diseases. Manually segmenting the corpus callosum from brain magnetic resonance (MR) image is time consuming, error prone, and operator dependent. In this paper, two semi-automatic segmentation methods are present: the active contour model-based approach and the active shape model-based approach. We tested these methods on an MR image of the human brain and found that the active contour approach had better segmentation accuracy but was slower than the active shape approach.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 1998년도 Proceedings of International Symposium on Remote Sensing
• /
• pp.246-251
• /
• 1998

This paper proposes a new building segmentation method from 1m resolution imagery using an Active Contour Model, known as "Balloons". The original balloons, which was designed by Cohen(Cohen, 1991) to extract features from medical images, are modified for building segmentation. The proposed method consists of two phases. Firstly, building boundaries are extracted by balloons with a given position on buildings from an operator. Since balloons actively adjust their shapes according to the boundaries, there is no more shape limitations on detecting buildings. Secondly, buildings are segmented by connecting the corners detected from the building boundaries, because most buildings, which are man-made objects, are effectively described by polygons. The test results show that most buildings are segmented efficiently and easily. The proposed method is new and timely as 1m resolution spaceborne imagery will be available in the very near future. The proposed method can be used fur operational building segmentation from such imagery.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2006년도 하계종합학술대회
• /
• pp.335-336
• /
• 2006

In this paper, we propose an active shape image segmentation method for three-dimensional(3-D) medical images using a generation method of the 3-D shape model. The proposed method generates the shape model using a distance transform and a tetrahedron method for landmarking. After generating the 3-D model, we extend the training and segmentation processes of 2-D active shape model(ASM) and improve the searching process. The proposed method provides comparative results to 2-D ASM, region-based or contour-based methods. Experimental results demonstrate that this algorithm is effective for a semi-automatic segmentation method of 3-D medical images.