• Journal of Korea Multimedia Society
• /
• v.8 no.8
• /
• pp.1069-1079
• /
• 2005

Though there have been many methods to detect features in spatial domain, in the case of a compressed image it has to be decoded, processed and encoded again. Alternatively, we can manipulate a compressed image directly in the Discrete Cosine Transform (DCT) domain that has been used for compressing videos or images in the standards like MPEG and JPEG. In our previous work we proposed a model-based discontinuity evaluation technique in the DCT domain that had problems in the rotated or non-ideal discontinuities. In this paper, we propose a fuzzy filtering technique that consists of height fuzzification, direction fuzzification, and forty filtering of discontinuities. The enhancement achieved by the fuzzy tittering includes the linking, thinning, and smoothing of discontinuities in the DCT domain. Although the detected discontinuities are rough in a low-resolution image for the size (8${\times}$8 pixels) of the DCT block, experimental results show that this technique is fast and stable to enhance the qualify of discontinuities.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.15 no.5
• /
• pp.64-74
• /
• 2019

Although a non-rigid registration has high demands in clinical practice, it has a high computational complexity and it is very difficult for ensuring the accuracy and robustness of registration. This study proposes a method of applying a non-rigid registration to 3D magnetic resonance images of brain in an unsupervised learning environment by using a deep-learning network. A feature vector between two images is produced through the network by receiving both images from two different patients as inputs and it transforms the target image to match the source image by creating a displacement vector field. The network is designed based on a U-Net shape so that feature vectors that consider all global and local differences between two images can be constructed when performing the registration. As a regularization term is added to a loss function, a transformation result similar to that of a real brain movement can be obtained after the application of trilinear interpolation. This method enables a non-rigid registration with a single-pass deformation by only receiving two arbitrary images as inputs through an unsupervised learning. Therefore, it can perform faster than other non-learning-based registration methods that require iterative optimization processes. Our experiment was performed with 3D magnetic resonance images of 50 human brains, and the measurement result of the dice similarity coefficient confirmed an approximately 16% similarity improvement by using our method after the registration. It also showed a similar performance compared with the non-learning-based method, with about 10,000 times speed increase. The proposed method can be used for non-rigid registration of various kinds of medical image data.

• Journal of the Korea Convergence Society
• /
• v.10 no.3
• /
• pp.39-47
• /
• 2019

In this paper, we proposed extracting modified Histogram of Oriented Gradients (HOG) features using background removal when tracking pedestrians in real time. HOG feature extraction has a problem of slow processing speed due to large computation amount. Background removal has been studied to improve computation reductions and tracking rate. Area removal was carried out using S and V channels in HSV color space to reduce feature extraction in unnecessary areas. The average S and V channels of the video were removed and the input video was totally dark, so that the object tracking may fail. Histogram equalization was performed to prevent this case. HOG features extracted from the removed region are reduced, and processing speed and tracking rates were improved by extracting clear HOG features. In this experiment, we experimented with videos with a large number of pedestrians or one pedestrian, complicated videos with backgrounds, and videos with severe tremors. Compared with the existing HOG-SVM method, the proposed method improved the processing speed by 41.84% and the error rate was reduced by 52.29%.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.26 no.6
• /
• pp.485-492
• /
• 2016

A role of the binarization is very important in separating the foreground and the background in the field of the computer vision. In this study, an adaptive fuzzy binarization is proposed. An ${\alpha}$-cut control ratio is obtained by the distribution of grey level of pixels in a sliding window, and binarization is performed using the value. To obtain the ${\alpha}$-cut, existing thresholding methods which execution speed is fast are used. The threshold values are set as the center of each membership function and the fuzzy intervals of the functions are specified with the distribution of grey level of the pixel. Then ${\alpha}$-control ratio is calculated using the specified function and binarization is performed according to the membership degree of the pixels. The experimental results show the proposed method can segment the foreground and the background well than existing binarization methods and decrease loss of the foreground.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.2
• /
• pp.88-95
• /
• 2011

Stereo matching is the research area that regarding the estimation of the distance between objects and camera using different view points and it still needs lot of improvements in aspects of speed and accuracy. This paper presents a fast stereo matching algorithm based on plane-converging belief propagation that uses message passing convergence in hierarchical belief propagation. Also, stereo matching technique is developed using GPU and it is available for real-time applications. The error rate of proposed Plane-converging Belief Propagation algorithm is similar to the conventional Hierarchical Belief Propagation algorithm, while speed-up factor reaches 2.7 times.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.11
• /
• pp.529-536
• /
• 2015

When rendering both three-dimensional objects and photo images together, the non-photorealistic rendering results are in visual discord since the two contents have their own independent color distributions. This paper proposes a non-photorealistic rendering technique which renders both three-dimensional objects and photo images such as cartoons and sketches. The proposed technique computes the color distribution property of the photo images and reduces the number of colors of both photo images and 3D objects. NPR is performed based on the reduced colormaps and edge features. To enhance the natural scene presentation, the image region segmentation process is preferred when extracting and applying colormaps. However, the image segmentation technique needs a lot of computational operations. It takes a long time for non-photorealistic rendering for large size frames. To speed up the time-consuming segmentation procedure, we use GPGPU for the parallel computing using the GPU. As a result, we significantly improve the execution speed of the algorithm.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.485-487
• /
• 2023

최근 코로나 19 방역지침 해제로 인한 대면적인 활동이 많아지면서 사람에 대한 서비스 제공이 중요한 이슈가 되었다. 하지만 사람들이 밀집되어있는 곳에서는 서비스가 원할하게 이루어지지 않는 경우가 대부분이다. 본 논문에서는 객체인식 알고리즘 기술인 Yolo와 OpenCv를 통해 카메라로 영상 속의 사람들을 인식하여 군집화 기술인 K-means 클러스터링을 이용해서 사람에 대한 군집화를 진행후 우선순위를 선정하고 좌표를 지정하여서 로봇이 군집의 좌표로 이동하여서 사람들에게 직접 접근하여 서비스를 제공할 수 있도록 하였다.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1408-1414
• /
• 2019

Deep artificial neural network with transfer learning is applied to compressed sensing cardiovascular MRI. Transfer learning is a method that utilizes structure, filter kernels, and weights of the network used in prior learning for current learning or application. The transfer learning is useful in accelerating learning speed, and in generalization of the neural network when learning data is limited. From a cardiac MRI experiment, with 8 healthy volunteers, the neural network with transfer learning was able to reduce learning time by a factor of more than five compared to that with standalone learning. Using test data set, reconstructed images with transfer learning showed lower normalized mean square error and better image quality compared to those without transfer learning.

• 한국전산유체공학회:학술대회논문집
• /
• 1998.11a
• /
• pp.13-18
• /
• 1998

The practical use of the particle image velocimetry(PIV), a whole-field velocity measurement method, requires the use of fast, reliable, computer-based methods for tracking velocity vectors. The full search block matching, the most widely studied and applied technique both in area of PIV and Image Coding and Compression, is computationally costly. Many less expensive alternatives have been proposed mostly in the area of Image Coding and Compression. Among others, TSS, NTSS, HPM are introduced for the past PIV analysis, and found to be successful. But, these algorithms are based on small dynamic range, 7 pixels/frame in maximum displacement. To analyze the images with large displacement, Even and Odd field image separation and a simple version of multi-resolution hierarchical procedures are introduced in this paper. Comparison with other algorithms are summarized. A Results of application to the turbulent backward step flow shows the improvement of new algorithm.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.1
• /
• pp.60-69
• /
• 1999

Motion compensation by affine transform has been proposed as a solution to the artifact problems in very low bit rate video coding and a HMA(Hexagoanl Matching Algorithm) has been proposed for refine motions estimation. When dividing images with an affine transform, as image objects do not necessarily conform to triangle patterns. In this paper we propose a method that first divides an image into triangular patches according to its edge information and then further divides the image into more detailed triangular patches where more complicated edge information occurs. We image propose a PMA(Polygona Matching Algorithm) for refine motion estimation because of the different triangle pattern types of neighboring blocks and its performance is compared with H.263.