• Proceedings of the Korea Contents Association Conference
• /
• 2012.05a
• /
• pp.53-54
• /
• 2012

본 논문에서는 Depth Camera를 통해 생성되는 깊이영상과 컬러영상을 통해 깊이정보와 영상정보를 추출하고, 추출된 깊이정보와 영상정보를 이용하여 집적영상을 OpenCL을 이용하여 실시간으로 생성하는 기법을 소개한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2012.07a
• /
• pp.127-130
• /
• 2012

본 논문에서는 자유시점 또는 오토스테레오스코픽 비디오나 홀로그래픽 비디오 서비스 들을 목표로 가상시점 영상을 빠르게 생성하는 방법을 제안한다. 이 방법에서 좌/우 깊이영상이 주어졌다고 가정하고, 두 깊이 영상을 이용하여 깊이 값과 변위를 계산 한 후 깊이 당 변위증분을 계산하여 사용한다. 계산된 깊이 값 당 변위증분과 해당 픽셀의 깊이정보를 이용하여 텍스쳐 이미지를 원하는 가상 시점으로 이동시킨다. 좌영상과 우영상을 원하는 가상시점으로 각각 생성한 후, 잡음과 disocclusion을 효과적으로 제거하기 위하여 두 영상을 결합한다. 그 다음 남은 홀을 효과적인 hole-filling으로 제거하여 가상 시점 영상을 생성한다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.6
• /
• pp.1311-1316
• /
• 2021

In this paper, we present a bifocal stereo camera system combining two cameras with different focal length cameras to generate stereoscopic image and their corresponding depth map. In order to obtain the depth data using the bifocal stereo camera system, we perform camera calibration to extract internal and external camera parameters for each camera. We calculate a common image plane and perform a image rectification for generating the depth map using camera parameters of bifocal stereo camera. Finally we use a SGM(Semi-global matching) algorithm to generate the depth map in this paper. The proposed bifocal stereo camera system can performs not only their own functions but also generates distance information about vehicles, pedestrians, and obstacles in the current driving environment. This made it possible to design safer autonomous vehicles.

• Journal of the Korea Computer Graphics Society
• /
• v.20 no.2
• /
• pp.1-11
• /
• 2014

The quality of 2D-to-3D conversion depends on the accuracy of the assigned depth to scene objects. Manual depth painting for given objects is labor intensive as each frame is painted. Specifically, a human is one of the most challenging objects for a high-quality conversion, as a human body is an articulated figure and has many degrees of freedom (DOF). In addition, various styles of clothes, accessories, and hair create a very complex silhouette around the 2D human object. We propose an efficient method to estimate visually pleasing depths of a human at every frame in a monocular video. First, a 3D template model is matched to a person in a monocular video with a small number of specified user correspondences. Our pose estimation with sequential joint angular constraints reproduces a various range of human motions (i.e., spine bending) by allowing the utilization of a fully skinned 3D model with a large number of joints and DOFs. The initial depth of the 2D object in the video is assigned from the matched results, and then propagated toward areas where the depth is missing to produce a complete depth map. For the effective handling of the complex silhouettes and appearances, we introduce a partial depth propagation method based on color segmentation to ensure the detail of the results. We compared the result and depth maps painted by experienced artists. The comparison shows that our method produces viable depth maps of humans in monocular videos efficiently.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.05a
• /
• pp.371-373
• /
• 2013

입체영상을 생성하기 위한 방법 중 깊이지도(depth-map)을 이용하는 방법이 있다. 깊이지도를 생성하기 위해서는 스테레오 시스템(stereo system)을 통한 두 대의 카메라 정보를 통하여 접근하는 것이 일반적인 방법으로, 시스템 구현, 연산량, 계산 오차 등의 이유로 상당히 제약사항이 많다. 본 논문에서는 모바일 시스템에서의 단일 카메라를 사용하여 영상 인식 및 부가정보를 통한 거리정보 생성에 대한 시스템을 제안한다.

• Journal of KIISE:Software and Applications
• /
• v.31 no.5
• /
• pp.617-624
• /
• 2004

This paper describes a new facial feature localization method that uses Adjacent Depth Differences(ADD) in 3D facial surface. In general, human recognize the extent of deepness or shallowness of region relatively, in depth, by comparing the neighboring depth information among regions of an object. The larger the depth difference between regions shows, the easier one can recognize each region. Using this principal, facial feature extraction will be easier, more reliable and speedy. 3D range images are used as input images. And ADD are obtained by differencing two range values, which are separated at a distance coordinate, both in horizontal and vertical directions. ADD and input image are analyzed to extract facial features, then localized a nose region, which is the most prominent feature in 3D facial surface, effectively and accurately.

• KIPS Transactions on Computer and Communication Systems
• /
• v.5 no.9
• /
• pp.237-246
• /
• 2016

DIBR (Depth Image Based Rendering) is a multimedia technology that generates the virtual multi-view images using a color image and a depth image, and it is used for creating glasses-less 3-dimensional display contents. This research describes the effect of depth accuracy about the objective quality of DIBR-based multi-view images. It first evaluated the minimum depth quantization bit that enables the minimum distortion so that people cannot recognize the quality degradation. It then presented the comparative analysis of non-uniform domain-division quantization versus regular linear quantization to find out how effectively express the accuracy of the depth information in same quantization levels according to scene properties.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.4
• /
• pp.1010-1019
• /
• 2013

Recently, research on hologram, which can realize perfect 3-dimensional imaging, becomes more active. But hologram has the defect that it costs a lot in generation. Thus, this paper proposes a method to generate a new digital hologram contents by manipulating the existing digital hologram contents. That is, this paper proposes the method to manipulating the digital hologram contents by manipulating and/or synthesizing the depth information to get a new digital hologram contents. The proposed methods have been experimented with various depth informations and digital hologram contents. For each kind of depth information, it has been manipulated for its position and distance. The result was converted to a digital hologram by the computer-generated hologram method and the resulting hologram was reconstructed.

• Journal of Digital Convergence
• /
• v.10 no.9
• /
• pp.351-356
• /
• 2012

This paper presents the 2D/3D conversion technique using depth map which is generated based on the haze and relative height cue. In cases that only the conventional haze information is used, errors in image without haze could be generated. To reduce this kind of errors, a new approach is proposed combining the haze information with depth map which is constructed based on the relative height cue. Also the gray scale image from Mean Shift Segmentation is combined with depth map of haze information to sharpen the object's contour lines, upgrading the quality of 3D image. Left and right view images are generated by DIBR(Depth Image Based Rendering) using input image and final depth map. The left and right images are used to generate red-cyan 3D image and the result is verified by measuring PSNR between the depth maps.

• Smart Media Journal
• /
• v.1 no.3
• /
• pp.29-35
• /
• 2012

A depth image-based rendering (DIBR) technique is one of the rendering processes of virtual views with a color image and the corresponding depth map. The most important issue of DIBR is that the virtual view has no information at newly exposed areas, so called dis-occlusion. In this paper, we propose an intermediate view generation algorithm using the Kinect depth camera that utilizes the infrared structured light. After we capture a color image and its corresponding depth map, we pre-process the depth map. The pre-processed depth map is warped to the virtual viewpoint and filtered by median filtering to reduce the truncation error. Then, the color image is back-projected to the virtual viewpoint using the warped depth map. In order to fill out the remaining holes caused by dis-occlusion, we perform a background-based image in-painting operation. Finally, we obtain the synthesized image without any dis-occlusion. From experimental results, we have shown that the proposed algorithm generated very natural images in real-time.