• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.2C
• /
• pp.16-24
• /
• 2005

In this paper, we proposed 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 can be synthesized which correspond to viewer's position by using geometrical transformation such as a rotation and a translation. The proposed face tracking technique gives a motion parallax cue by different viewpoints and view angles. In the proposed algorithm, tracking of viewer's dominant face initially established from camera by using statistical characteristics of face colors and deformable templates is done. As a result, we can provide motion parallax cue by detecting viewer's dominant face area and tracking it even under a heterogeneous background and can successfully display the synthesized sequences.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.21 no.3
• /
• pp.245-252
• /
• 2010

This paper proposes an improved filter synthesis method for dual-band filters based on coupled lines loaded with shunt stubs. The conventional design method relies on equality of the response of the conventional single-band filter and its dual-band counterpart at two frequencies. On the other hand, the proposed method is based on equivalence of the propagation constants and the image impedances of the two filters. Therefore, the proposed method does not suffer from bandwidth reduction phenomena, which requires complex compensating process in the conventional method. Also, the alternative design parameters provided by the demonstrated method enable to construct dual-band filters with more flexibility. For experimental verification, a Chebyshev type dual-band filter with center frequencies at 1 GHz and 3.5 GHz is designed and fabricated. The measured results are in excellent agreement with the full-wave simulated results.

• Imaging Science in Dentistry
• /
• v.52 no.4
• /
• pp.393-398
• /
• 2022

Purpose: This study proposed a generative adversarial network (GAN) model for T2-weighted image (WI) synthesis from proton density (PD)-WI in a temporomandibular joint(TMJ) magnetic resonance imaging (MRI) protocol. Materials and Methods: From January to November 2019, MRI scans for TMJ were reviewed and 308 imaging sets were collected. For training, 277 pairs of PD- and T2-WI sagittal TMJ images were used. Transfer learning of the pix2pix GAN model was utilized to generate T2-WI from PD-WI. Model performance was evaluated with the structural similarity index map (SSIM) and peak signal-to-noise ratio (PSNR) indices for 31 predicted T2-WI (pT2). The disc position was clinically diagnosed as anterior disc displacement with or without reduction, and joint effusion as present or absent. The true T2-WI-based diagnosis was regarded as the gold standard, to which pT2-based diagnoses were compared using Cohen's ĸ coefficient. Results: The mean SSIM and PSNR values were 0.4781(±0.0522) and 21.30(±1.51) dB, respectively. The pT2 protocol showed almost perfect agreement(ĸ=0.81) with the gold standard for disc position. The number of discordant cases was higher for normal disc position (17%) than for anterior displacement with reduction (2%) or without reduction (10%). The effusion diagnosis also showed almost perfect agreement(ĸ=0.88), with higher concordance for the presence (85%) than for the absence (77%) of effusion. Conclusion: The application of pT2 images for a TMJ MRI protocol useful for diagnosis, although the image quality of pT2 was not fully satisfactory. Further research is expected to enhance pT2 quality.

• Journal of the Microelectronics and Packaging Society
• /
• v.30 no.4
• /
• pp.69-78
• /
• 2023

With the rapid growth of artificial intelligence, the demand for semiconductors is enormously increasing everywhere. To ensure the manufacturing quality and quantity simultaneously, the importance of automatic defect detection during the packaging process has been re-visited by adapting various deep learning-based methodologies into automatic packaging defect inspection. Deep learning (DL) models require a large amount of data for training, but due to the nature of the semiconductor industry where security is important, sharing and labeling of relevant data is challenging, making it difficult for model training. In this study, we propose a new framework for securing sufficient data for DL models with fewer computing resources through a divide-and-conquer approach. The proposed method divides high-resolution images into pre-defined sub-regions and assigns conditional labels to each region, then trains individual sub-regions and boundaries with boundary loss inducing the globally coherent and seamless images. Afterwards, full-size image is reconstructed by combining divided sub-regions. The experimental results show that the images obtained through this research have high efficiency, consistency, quality, and generality.

• Journal of KIISE
• /
• v.44 no.2
• /
• pp.186-194
• /
• 2017

Depth-image-based rendering is a technique that can be applied in a variety of 3D-display systems. It generates the images that have been captured from virtual viewpoints by using a depth map. However, disoccluded hole-filling problems remain a challenging issue, as a newly exposed area appears in the virtual view. Image inpainting is a popular approach for the filling of the hole region. This paper presents a robust hole-filling method that reduces the error and generates a high quality-virtual view. First, the adaptive-patch size is decided using the color and depth information. Also, a partial filling method for which the patch similarity is used is proposed. These efforts reduce the error occurrence and the propagation. The experiment results show that the proposed method synthesizes the virtual view with a higher visual comfort compared with the existing methods.

• Journal of Internet of Things and Convergence
• /
• v.4 no.2
• /
• pp.1-6
• /
• 2018

Techniques for generating new sample data from higher dimensional data such as images have been utilized variously for speech synthesis, image conversion and image restoration. This paper adopts Progressive Growing of Generative Adversarial Networks(PG-GANs) as an implementation model to generate high-resolution images and to enhance variation of the generated images, and applied it to fashion image data. PG-GANs allows the generator and discriminator to progressively learn at the same time, continuously adding new layers from low-resolution images to result high-resolution images. We also proposed a Mini-batch Discrimination method to increase the diversity of generated data, and proposed a Sliced Wasserstein Distance(SWD) evaluation method instead of the existing MS-SSIM to evaluate the GAN model.

• The Journal of the Korea Contents Association
• /
• v.14 no.12
• /
• pp.555-564
• /
• 2014

The boundary noise in image syntheses using DIBR consists of noisy pixels that are separated from foreground objects into background region. It is generated mainly by edge misalignment between the reference image and depth map or blurred edge in the reference image. Since hole areas are generally filled with neighboring pixels, boundary noise adjacent to the hole is the main cause of quality degradation in synthesized images. To solve this problem, a new boundary noise removal algorithm using a preprocessing of the depth map is proposed in this paper. The most common way to eliminate boundary noise caused by boundary misalignment is to modify depth map so that the boundary of the depth map can be matched to that of the reference image. Most conventional methods, however, show poor performances of boundary detection especially in blurred edge, because they are based on a simple boundary search algorithm which exploits signal gradient. In the proposed method, a two-step hierarchical approach for boundary detection is adopted which enables effective boundary detection between the transition and background regions. Experimental results show that the proposed method outperforms conventional ones subjectively and objectively.

• Journal of Broadcast Engineering
• /
• v.19 no.3
• /
• pp.316-328
• /
• 2014

In this paper, we parallelize the popular feature detection algorithms, i.e. SIFT and SURF, and its application to fast panoramic image generation on the latest embedded GPU. Parallelized algorithms are implemented using recently developed OpenCL as the embedded GPGPU software platform. We compare the implementation efficiency and speed performance of conventional OpenGL Shading Language and OpenCL. Experimental result shows that implementation on OpenCL has comparable performance with GLSL. Compared with the performance on the embedded CPU in the same application processor, the embedded GPU runs 3~4 times faster. As an example of using feature extraction, panorama image synthesis is performed on embedded GPU by applying image matching using detected features.

• Journal of Broadcast Engineering
• /
• v.22 no.3
• /
• pp.339-349
• /
• 2017

In this paper, we propose a technique for smoothly replacing a signboard image in a video clip with another signboard image and an application for performing it. In order to replace the signboard area in the image, it must be tracked and the shape of the signboard changed. Also, when an obstruction occurs due to a front object, the obscured area is extracted so that only the signboard area can be replaced, and color correction is performed for natural synthesis. The proposed application receives the information required by the algorithm from the user and explains the required functions, thereby disclosing the entire process in which the algorithm is actually implemented and tested. This can be used for a variety of media, and can be used as a means to remove the indirect advertisement that is often seen, and it can also be used as a new advertisement model in which another advertisement image is inserted into the image to be advertised.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.12
• /
• pp.49-60
• /
• 2012

In this paper, we propose a hardware design for correction of barrel distortion using the nearest coordinates in the corrected image. Because it applies the nearest distance on corrected image rather than adjacent distance on distorted image, the picture quality is improved by the image whole area, solve the staircase phenomenon in the exterior area. But, because of additional arithmetic operation using design of bilinear interpolation, required arithmetic operation is increased. Look up table(LUT) structure is proposed in order to solve this, coordinate rotation digital computer(CORDIC) algorithm is applied. The results of the synthesis using Design compiler, the design of implementing all processes of the interpolation method with the hardware is higher than the previous design about the throughput, In case of the rear camera, the design of using LUT and hardware together can reduce the size than the design of implementing all processes with the hardware.