• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.7
• /
• pp.2328-2344
• /
• 2022

2D human pose estimation still faces difficulty in low-resolution images. Most existing top-down approaches scale up the target human bonding box images to the large size and insert the scaled image into the network. Due to up-sampling, artifacts occur in the low-resolution target images, and the degraded images adversely affect the accurate estimation of the joint positions. To address this issue, we propose a multi-resolution input feature fusion network for human pose estimation. Specifically, the bounding box image of the target human is rescaled to multiple input images of various sizes, and the features extracted from the multiple images are fused in the network. Moreover, we introduce a guiding channel which induces the multi-resolution input features to alternatively affect the network according to the resolution of the target image. We conduct experiments on MS COCO dataset which is a representative dataset for 2D human pose estimation, where our method achieves superior performance compared to the strong baseline HRNet and the previous state-of-the-art methods.

• Journal of Korea Multimedia Society
• /
• v.24 no.8
• /
• pp.979-987
• /
• 2021

Multiple Sclerosis (MS) can be early diagnosed by detecting lesions in brain magnetic resonance images (MRI). Unsupervised anomaly detection methods based on autoencoder have been recently proposed for automated detection of MS lesions. However, these autoencoder-based methods were developed only for 2D images (e.g. 2D cross-sectional slices) of MRI, so do not utilize the full 3D information of MRI. In this paper, therefore, we propose a novel 3D autoencoder-based framework for detection of the lesion volume of MS in MRI. We first define a 3D convolutional neural network (CNN) for full MRI volumes, and build each encoder and decoder layer of the 3D autoencoder based on 3D CNN. We also add a skip connection between the encoder and decoder layer for effective data reconstruction. In the experimental results, we compare the 3D autoencoder-based method with the 2D autoencoder models using the training datasets of 80 healthy subjects from the Human Connectome Project (HCP) and the testing datasets of 25 MS patients from the Longitudinal multiple sclerosis lesion segmentation challenge, and show that the proposed method achieves superior performance in prediction of MS lesion by up to 15%.

• Journal of Korea Game Society
• /
• v.3 no.2
• /
• pp.30-34
• /
• 2003

Image-based rendering is an approach to generate realistic images in real-time without modeling explicit 3D geometry, Especially, TIP(Tour Into the Picture) is preferred for its simplicity in constructing 3D background scene. However, TP has a limitation that a viewpoint cannot go far from the origin of the TIP for the lack of geometrical information. in this paper, we propose a method to interpolating the TIP images to generate smooth and realistic navigation. We construct multiple TIP models in a wide area of the virtual environment. Then we interpolate foreground objects and background object respectively to generate smooth navigation results.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.49 no.12
• /
• pp.683-689
• /
• 2000

Two-dimensional(2-D) fiber-optic code-division multiple-access(FO-CDMA) system utilizes the optical orthogonal signature pattern code(OOSPC) to encode and decode 2-D data. Encoded 2-D data are spatially multiplexed and transmitted through an image fiber and receiver recovers the intended data by means of thresholding process. OOSPC's construction methods based on expansion of the optical orthogonal code, which is used in one-dimensional(1-D) FO-CDMA system, are introduced. Each OOSPC's performances are compared by using the bit error rate(BER) of interfering OOSPC's of other users. From the results we verify that a balanced incomplete block design(BIBD) construction has the best performance among other mehtods. We also propose a decomposed bit-plane method for parallel transmission and recovery of 256 gray-scale images using OOSPC's constructed by the BIBD method. The simulation result encourages the feasibility of parallel transmission and recovery of multiuser's images.

• The KIPS Transactions:PartA
• /
• v.9A no.2
• /
• pp.217-224
• /
• 2002

Many systems handle 3D-image were used. High-performance computer systems and techniques of compressing images to handle 3D-image were used. But there will be cost Problems, if GIS system is implemented, using the high-performance system. And if GIS system is implemented, using the techniques of compressing images, there will be some loss of a image. It will take a long processing time to handle 3D-images using a general PC because the size of 3D-image files are very huge. The parallel algorithm presented in the paper can improve speed to handle 3D-image using parallel computer system. The system uses the method of displacing images from nodes to screens, dividing a 3D-image into multiple sub images on multiple nodes. The performance of the presented algorithm showers improving speed by experiments.

• Journal of information and communication convergence engineering
• /
• v.7 no.1
• /
• pp.1-6
• /
• 2009

In order to obtain a 3-D urban model, an abstraction of the surface model is required. This paper describes works on the 3D reconstruction and modeling by the grouping 3D line segments extracted from the stereo matching of edges, which is derived from multiple images. The grouping is achieved by conditions of degrees and distances between lines. Building objects are determined by the junction combinations of the grouped line segments. The proposed algorithm demonstrates effective results of 3D reconstruction of buildings with 2D aerial images.

• 한국정보디스플레이학회:학술대회논문집
• /
• 2002.08a
• /
• pp.289-292
• /
• 2002

A new technique to construct an auto-stereoscopic display that offers massive horizontal parallax images is proposed Multiple telecetnric imaging systems are arranged in a modified 2D array. The horizontal parallax images displayed by LCD panels are imaged to be superimposed on a 3D screen. All parallax images are displayed in the different horizontal directions because all imaging systems have different horizontal positions. The difference of the vertical display directions due to the imaging system's vertical positions is canceled by a vertical diffuser placed at the 3D screen. Observers can percept 3D images with the binocular disparity, the vergence, and the smooth motion parallax. In addition, the accommodation function may also work because a number of parallax images are displayed with a very small angle interval in the horizontal direction. A prototype 3D display including 64 color LCD panels was constructed.

• Journal of Korea Multimedia Society
• /
• v.18 no.6
• /
• pp.701-709
• /
• 2015

Image stitching method is a technique for obtaining an high-resolution image by combining two or more images. In X-ray image for clinical diagnosis, the size of the imaging region taken by one shot is limited due to the field-of-view of the equipment. Therefore, in order to obtain a high-resolution image including large regions such as a whole body, the synthesis of multiple X-ray images is required. In this paper, we propose a rapid stitching method of digital X-ray images using template-based registration. The proposed algorithm use principal component analysis(PCA) and k-nearest neighborhood(k-NN) to determine the location of input images before performing a template-based matching. After detecting the overlapping position using template-based matching, we synthesize input images by alpha blending. To improve the computational efficiency, reduced images are used for PCA and k-NN analysis. Experimental results showed that our method was more accurate comparing with the previous method with the improvement of the registration speed. Our stitching method could be usefully applied into the stitching of 2D or 3D multiple images.

• Computers and Concrete
• /
• v.17 no.6
• /
• pp.831-848
• /
• 2016

This study aims to characterize the multiple cracking behavior of HTPP-ECC (High tenacity polypropylene fiber reinforced engineered cementitious composites) by Digital Image Correlation (DIC) Method. Digital images have been captured from a dogbone shaped HTPP-ECC specimen exhibiting 3.1% tensile ductility under loading. Images analyzed by VIC-2D software and ${\varepsilon}_{xx}$ strain maps have been obtained. Crack widths were computed from the ${\varepsilon}_{xx}$ strain maps and crack width distributions were determined throughout the specimen. The strain values from real LVDTs were also compared with virtual LVDTs digitally attached on digital images. Results confirmed that it is possible to accurately monitor the initiation and propagation of any single crack or multiple cracks by DIC at the whole interval of testing. Although the analysis require some post-processing operations, DIC based crack analysis methodology can be used as a promising and versatile tool for quality control of HTPP-ECC and other strain hardening composites.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.3 s.303
• /
• pp.41-50
• /
• 2005

A novel image mosaicking technique suitable for 3D urban visualization is proposed. It is not effective to apply 2D image mosaicking techniques for urban visualization when, for example, one is filming a sequence of images from a side-looking video camera along a road in an urban area. The proposed method presents the roadside scene captured by a side-looking video camera as a continuous set of textured planar faces, which are termed 'multiple planes' in this paper. The exterior parameters of each frame are first calculated through automatically selected matching feature points. The matching feature points are also used to estimate a plane approximation of the scene geometry for each frame. These planes are concatenated to create an approximate model on which images are back-projected as textures. Here, we demonstrate algorithm that creates efficient image mosaics in 3D space from a sequence of real images.