• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.3
• /
• pp.129-135
• /
• 2011

In this paper, we propose the algorithm using optical flow and machine learning-based segmentation for the 3D conversion of 2D video. For the segmentation allowing the successful 3D conversion, we design a new energy function, where color/texture features are included through machine learning method and the optical flow is also introduced in order to focus on the regions with the motion. The depth map are then calculated according to the optical flow of segmented regions, and left/right images for the 3D conversion are produced. Experiment on various video shows that the proposed method yields the reliable segmentation result and depth map for the 3D conversion of 2D video.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2016.06a
• /
• pp.1-4
• /
• 2016

효율적으로 깊이지도를 획득하기 위해 다양한 방법의 지역 기반스테레오 매칭 방법이 사용된다. 일반적인 지역기반 스테레오 매칭에 사용되는 비용값 계산 방법을 통해 깊이지도를 생성하게 되면 객체의 경계 영역이 무너지거나, 유사한 텍스쳐 정보가 연속적으로 나타나는 영역에서 부정확한 깊이값을 얻는 문제가 발생한다. 본 논문에서는 깊이지도의 정확성을 높이기 위해 2가지 단계를 거쳐 최종 깊이지도를 생성한다. 처음으로, 일반적으로 사용하는 지역기반 스테레오 매칭 비용 함수와 입력 영상의 기울기를 고려한 초기 비용값을 가이드 필터를 이용하여 최적의 비용값을 찾아 초기 변위지도를 생성한다. 스테레오매칭을 수행할 경우, 시점의 차이로 인해 보이지 않는 영역에서 정확한 변위값을 찾지 못하는 문제가 발생한다. 이러한 문제를 해결하기 위해 좌영상과 우영상을 기반으로 획득한 변위지도를 사용하여 교차검사를 함으로써 폐색영역을 찾아낸다. 폐색 영역을 이웃한 화소의 값을 사용하여 채울 경우 실선과 같은 오류가 결과 영상에 나타나게 된다. 이러한 오류 영역을 제거하기 위해 마지막으로 가중치를 적용한 중간값 필터를 적용한다. 실험 결과 제안한 방법을 사용하여 획득한 깊이지도가 기존의 방법보다 정확한 깊이값을 얻는 것을 확인할 수 있었다.

• Proceedings of the KIEE Conference
• /
• 2008.10b
• /
• pp.268-269
• /
• 2008

Moving object detection is a fundamental technique in visual surveillance. Robust technique to enhance performance of moving object detection is required for several bad conditions in real external circumtance. In case of sudden illumination change in outdoor condition, many objects are determined as moving object though they are not really moving, but just their illumination changes. This makes the detection result untrustworthy. In this paper, robust moving object detection to sudden illumination change using gaussian mixture background model and new texture information using background from the weighted sum of recent images is proposed.

• Journal of Korea Multimedia Society
• /
• v.19 no.6
• /
• pp.979-989
• /
• 2016

Traversable region detection is an essential step for advanced driver assistance systems and self-driving car systems, and it has been conducted by detecting lanes from input images. The performance can be unreliable, however, when the light condition is poor or there exist no lanes on the roads. To solve this problem, this paper proposes an algorithm which utilizes the information about the number of lanes and texture analysis. The proposed algorithm first specifies road region candidates by utilizing the number of lanes information. Among road region candidates, the road region is determined as the region in which texture is homogeneous and texture discontinuities occur around its boundaries. Traversable region is finally detected by dividing the estimated road region with the number of lanes information. This paper combines the proposed algorithm with a lane detection-based method to construct a system, and simulation results show that the system detects traversable region even on the road with poor light conditions or no lanes.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.9
• /
• pp.76-90
• /
• 1999

In this paper, a color image retrieval algorithm is proposed based on color histogram and color texture. The representative color vectors of a color image are made from k-means clustering of its color histogram, and color texture is generated by centering around the color of pixels with its color vector. Thus the color texture means texture properties emphasized by its color histogram, and it is analyzed by Gaussian Markov Random Field (GMRF) model. The proposed algorithm can work efficiently because it does not require any low level image processing such as segmentation or edge detection, so it outperforms the traditional algorithms which use color histogram only or texture properties come from image intensity.

• Journal of the Korea Computer Graphics Society
• /
• v.16 no.4
• /
• pp.11-16
• /
• 2010

We present an efficient algorithm for rendering sweep surfaces using programmable graphics hardware. A sweep surface can be represented by a cross-section curve undergoing a spline motion. This representation has a simple matrix-vector multiplication structure that can easily be adapted to programmable graphics hardware. The data for the motion and cross-section curves are stored in texture memory. The vertex processor considers a pair of surface parameters as a vertex and evaluates its coordinates and normal vector with a single matrix multiplication. Using the GPU in this way is between 10 and 40 times as fast as CPU-based rendering.

• Journal of KIISE:Computer Systems and Theory
• /
• v.37 no.2
• /
• pp.110-125
• /
• 2010

We present a novel scheme for generating line drawings from 2D images, aiming to facilitate effective visual communication. In contrast to conventional edge detectors, our technique imitates the human line drawing process to generate lines effectively and intuitively. Our technique consists of three parts: line extraction, line rendering, and user guidance. In line extraction, we extract lines by estimating a likelihood function to effectively find the genuine shape boundaries. In line rendering, we consider the feature scale and the blurriness of lines with which the detail and the focus-level of lines are controlled. We also employ stroke textures to provide a variety of illustration styles. User guidance is allowed to modify the shapes and positions of lines interactively, where immediate response is provided by GPU implementation of most line extraction operations. Experimental results demonstrate that our technique generates various kinds of line drawings from 2D images enabled by the control over detail, focus, and style.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.7
• /
• pp.1763-1772
• /
• 1996

The conventional training set selection methods for the satellite image classification usually depend on the manual selection using data from the direct measurements of the ground or the ground map. However this task takes much time and cost, and some feature values vary in wide ranges even if they are in the same class. Such feature values can increase the robustness of the neural net but learning time becomes longer. In this paper,we propose anew training set selection algorithm using a rule-based method. By the technique proposed, the SPOT multispectral Imagery is classified in 3 bands, and the pixels which satisfy the rule are employed as the training sets for the neutralist classifier. The experimental results show faster initial convergence and almost the same or better classification accuracy. We also showed an improvement of the classification accuracy by using texture features and NDV1.

• The KIPS Transactions:PartA
• /
• v.14A no.7
• /
• pp.435-442
• /
• 2007

This paper is concerned with the mesh segmentation problem that can be applied to diverse applications such as texture mapping, simplification, morphing, compression, and shape matching for 3D mesh models. The mesh segmentation is the process of dividing a given mesh into the disjoint set of sub-meshes. We propose a method for segmenting meshes by simultaneously reflecting global and local geometric characteristics of the meshes. First, we extract sharp vertices over mesh vertices by interpreting the curvatures and convexity of a given mesh, which are respectively contained in the local and global geometric characteristics of the mesh. Next, we partition the sharp vertices into the $\kappa$ number of clusters by adopting the $\kappa$-means clustering method [29] based on the Euclidean distances between all pairs of the sharp vertices. Other vertices excluding the sharp vertices are merged into the nearest clusters by Euclidean distances. Also we implement the proposed method and visualize its experimental results on several 3D mesh models.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.6
• /
• pp.89-101
• /
• 2014

In this paper, we proposed a new local micro pattern, Signed Local Directional Pattern(SLDP). SLDP uses information of edges to represent the face's texture. This can produce a more discriminating and efficient code than other state-of-the-art methods. Each micro pattern of SLDP is encoded by sign and its major directions in which maximum edge responses exist-which allows it to distinguish among similar edge patterns that have different intensity transitions. In this paper, we divide the face image into several regions, each of which is used to calculate the distributions of the SLDP codes. Each distribution represents features of the region and these features are concatenated into a feature vector. We carried out facial expression recognition with feature vectors and SVM(Support Vector Machine) on Cohn-Kanade and JAFFE databases. SLDP shows better classification accuracy than other existing methods.