• Journal of IKEEE
• /
• v.28 no.1
• /
• pp.65-71
• /
• 2024

This paper proposes a neural network that can reduce the number of parameters while reducing matching errors in occluded regions to increase the accuracy of depth maps in stereo matching. Stereo matching-based object recognition is utilized in many fields to more accurately recognize situations using images. When there are many objects in a complex image, an occluded area is generated due to overlap between objects and occlusion by background, thereby lowering the accuracy of the depth map. To solve this problem, existing research methods that create context information and combine it with the cost volume or RoIselect in the occluded area increase the complexity of neural networks, making it difficult to learn and expensive to implement. In this paper, we create a depthwise seperable neural network that enhances regional feature extraction before cost volume generation, reducing the number of parameters and proposing a neural network that is robust to occlusion errors. Compared to PSMNet, the proposed neural network reduced the number of parameters by 30%, improving 5.3% in color error and 3.6% in test loss.

• Journal of Broadcast Engineering
• /
• v.2 no.2
• /
• pp.216-224
• /
• 1997

In this paper, we propose a depth information extraction method for intermediate image synthesis. As stereo matching method, MBS(Multiple-Baseline Stereo) method has been proposed, in which the matching accuracy increases by using the multiple camera, but there are some inherent problems such as computational complexity, boundary overreach(BO) in depth map, and occlusion. So, we propose the modified version of MBS so called Multi-Resolution MBS(MR-MBS). Moreover, we also propose an adaptive occlusion area processing technique to improve the accuracy of the depth information in occlusion area.

• IEIE Transactions on Smart Processing and Computing
• /
• v.2 no.1
• /
• pp.27-35
• /
• 2013

This paper proposes a face tracking algorithm for a viewpoint adaptive multi-view synthesis system. The original scene captured by a depth camera contains a texture image and 8 bit gray-scale depth map. From this original image, multi-view images that correspond to the viewer's position can be synthesized using geometrical transformations, such as rotation and translation. The proposed face tracking technique gives a motion parallax cue by different viewpoints and view angles. In the proposed algorithm, the viewer's dominant face, which is established initially from a camera, can be tracked using the statistical characteristics of face colors and deformable templates. As a result, a motion parallax cue can be provided by detecting the viewer's dominant face area and tracking it, even under a heterogeneous background, and synthesized sequences can be displayed successfully.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.6
• /
• pp.1194-1199
• /
• 2009

Stereo vision system is useful method for inferring 3D depth information from two or more images. So it has been the focus of attention in this field for a long time. Stereo matching is the process of finding correspondence points in two or more images. A central problem in a stereo matching is that it is difficult to satisfy both the computation time problem and accuracy at the same time. To resolve this problem, we proposed a new stereo matching technique using weighted cost aggregation. To begin with, we extract the weight in given stereo images based on features. We compute the costs of the pixels in a given window using correlation of weighted color and brightness information. Then, we match pixels in a given window between the reference and target images of a stereo pair. To demonstrate the effectiveness of the algorithm, we provide experimental data from several synthetic and real scenes. The experimental results show the improved accuracy of the proposed method.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2009.05a
• /
• pp.396-399
• /
• 2009

Stereo vision system is useful method for inferring 3D depth information from two or more images. So it has been the focus of attention in this field for a long time. Stereo matching is the process of finding correspondence points in two or more images. A central problem in a stereo matching is that it is difficult to satisfy both the computation time problem and accuracy at the same time. To resolve this problem, we proposed a new stereo matching technique using weighted cost aggregation. To begin with, we extract the weight in given stereo images based on features. We compute the costs of the pixels in a given window using correlation of weighted color, brightness and distance information. Then, we match pixels in a given window between the reference and target images of a stereo pair. To demonstrate the effectiveness of the algorithm, we provide experimental data from several synthetic and real scenes. The experimental results show the improved accuracy of the proposed method.

• Journal of Digital Convergence
• /
• v.11 no.12
• /
• pp.345-352
• /
• 2013

In this paper, a depth estimate method is proposed using GPU(Graphics Processing Unit) in monocular image. a monocular image is a 2D image with missing 3D depth information due to the camera projection and we used a monocular cue to recover the lost depth information by the projection present. The proposed algorithm uses an energy function which takes a variety of cues to create a more generalized and reliable depth map. But, a processing time is late because energy function is defined from the various monocular cues. Therefore, we propose a depth estimate method using GPGPU(General Purpose Graphics Processing Unit). The objective effectiveness of the algorithm is shown using PSNR(Peak Signal to Noise Ratio), a processing time is decrease by 61.22%.

• 한국HCI학회:학술대회논문집
• /
• 2007.02a
• /
• pp.1031-1036
• /
• 2007

최근 가상현실 기술의 주요 연구 동향으로 몰입감을 증가시키는 실감공간 구현구술이 주목 받고 있다. 실감공간 기술이란 서로 다른 공간에 떨어져 있는 사용자가 같은 공간에 있는 효과를 구현하는 기술이다. 본 논문에서는 특히 상호간의 주변 환경을 일치시키는 기술에 중점을 두고, 실시간으로 두 공간의 조명정보를 일치시키는 기술로서 2가지 핵심 내용을 소개한다. 첫째는 비주얼 헐 데이터를 기반으로 고속으로 노말벡터를 추출하는 방법이고, 둘째는 사용자 주변 조명 환경 정보를 반영하는 라이팅 방법이다. 본 논문에서 수행한 첫번째 방법은 비주얼 헐 데이터의 depth존재영역에서 노말맵을 계산하도록 하고, 노말맵을 계산할 때 주변 폴리곤들 기하학적 변화가 심할수록 노말맵 계산에 사용하는 주변 벡터의 선태을 늘리거나 줄이는 방식으로, 불필요한 계산량을 감소시켰다. 본 논문에서 수행한 두번째 방법에서는 주변 조명 정보에서 빛의 세기와 라이팅을 반영할 객체의 반사율의 특성을 고려하여 라이팅에 사용할 광원을 선택적으로 반영하여 불필요한 연산량을 감소시켰다. 종래의 영상기반 라이팅 기술이 사전에 촬영된 영상을 사용하거나 정지영상에 적용되는 연구를 한 반면에 본 논문은 실시간에서 라이팅을 구현하기 위한 시도로서 고속 라이팅 연산 기법을 제시하고 있다. 본 연구의 결과를 이용하면 영상기반 라이팅 연구의 실제적이고도 폭넓은 적용이 가능할 것으로 사료되며 고화질의 콘텐츠 양산에도 기여할 것으로 사료된다.

• Journal of The Geomorphological Association of Korea
• /
• v.26 no.2
• /
• pp.1-14
• /
• 2019

This study analyzed the relativity between block stream and talus distributions by employing a likelihood ratio approach. Possible distribution sites for each debris slope landform were extracted by applying a spatial integration model, in which we combined fuzzy set model, Bayesian predictive model, and logistic regression model. Moreover, to verify model performance, a success rate curve was prepared by cross-validation. The results showed that elevation, slope, curvature, topographic wetness index, geology, soil drainage, and soil depth were closely related to the debris slope landform sites. In addition, all spatial integration models displayed an accuracy of over 90%. The accuracy of the distribution potential area map of the block stream was highest in the logistic regression model (93.79%). Eventually, the accuracy of the distribution potential area map of the talus was also highest in the logistic regression model (97.02%). We expect that the present results will provide essential data and propose methodologies to improve the performance of efficient and systematic micro-landform studies. Moreover, our research will potentially help to enhance field research and topographic resource management.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.9
• /
• pp.2037-2042
• /
• 2012

Synthetic aperture integral imaging is one of promising 3D imaging techniques to capture the high-resolution elemental images using multiple cameras. In this paper, we propose a method of displaying 3D images in space using the synthetic aperture integral imaging technique. Since the elemental images captured from SAII cannot be directly used to display 3D images in an integral imaging display system, we first extract the depth map from elemental images and then transform them to novel elemental images for 3D image display. The newly generated elemental images are displayed on a display panel to generate 3D images in space. To show the usefulness of the proposed method, we carry out the preliminary experiments using a 3D toy object and present the experimental results.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.3
• /
• pp.1086-1103
• /
• 2020

The automatic identification and classification of image-based weld defects is a difficult task due to the complex texture of the X-ray images of the weld defect. Several depth learning methods for automatically identifying welds were proposed and tested. In this work, four different depth convolutional neural networks were evaluated and compared on the 1631 image set. The concavity, undercut, bar defects, circular defects, unfused defects and incomplete penetration in the weld image 6 different types of defects are classified. Another contribution of this paper is to train a CNN model "RayNet" for the dataset from scratch. In the experiment part, the parameters of convolution operation are compared and analyzed, in which the experimental part performs a comparative analysis of various parameters in the convolution operation, compares the size of the input image, gives the classification results for each defect, and finally shows the partial feature map during feature extraction with the classification accuracy reaching 96.5%, which is 6.6% higher than the classification accuracy of other existing fine-tuned models, and even improves the classification accuracy compared with the traditional image processing methods, and also proves that the model trained from scratch also has a good performance on small-scale data sets. Our proposed method can assist the evaluators in classifying pipeline welding defects.