• Journal of Korea Game Society
• /
• v.20 no.6
• /
• pp.43-52
• /
• 2020

This paper presents a fast pattern tracking method using location prediction in cubemap video for 360-degree VR. A spherical cubemap frame has six face textures and searching a pattern is much slower than a flat image. To overcome the limitation, we propose a method of predicting the location of target pattern using Kalman filter and reducing the search area by considering only textures of predicted location. The experimental results showed that the proposed system is much faster than the previous method of searching all six faces and also gives accurate pattern tracking performance.

• Journal of Korea Game Society
• /
• v.17 no.6
• /
• pp.153-164
• /
• 2017

In spherical panoramic images, SURF, a feature description method for planar patterns, does not work correctly due to heavy spherical distortion. Since a plane pattern is distorted in a spherical image, the pattern search and replacement in a spherical panoramic image should be treated differently from the case of the planar image. This paper proposes a planar texture replacement method, which transforms a spherical panoramic image into a cubemap panoramic image, searches a pattern using SURF, replaces a plane pattern, and then converts it into a spherical panoramic image.

• Journal of the Korea Computer Graphics Society
• /
• v.23 no.3
• /
• pp.65-75
• /
• 2017

As a 360-degree image carries information of all directions, it often has too much information. Moreover, in order to investigate a 360-degree image on a 2D display, a user has to either click and drag the image with a mouse, or project it to a 2D panorama image, which inevitably introduces severe distortions. In consequence, investigating a 360-degree image and finding an object of interest in such a 360-degree image could be a tedious task. To resolve this issue, this paper proposes a method to find a region of interest and produces a 2D naturally looking image from a given 360-degree image that best matches a description given by a user in a natural language sentence. Our method also considers photo composition so that the resulting image is aesthetically pleasing. Our method first converts a 360-degree image to a 2D cubemap. As objects in a 360-degree image may appear distorted or split into multiple pieces in a typical cubemap, leading to failure of detection of such objects, we introduce a modified cubemap. Then our method applies a Long Short Term Memory (LSTM) network based object detection method to find a region of interest with a given natural language sentence. Finally, our method produces an image that contains the detected region, and also has aesthetically pleasing composition.

• Journal of KIISE:Databases
• /
• v.30 no.3
• /
• pp.225-236
• /
• 2003

MOLAP systems store data in a multidimensional away called a 'cube' and access them using way indexes. When a cube is placed into disk, it can be Partitioned into a set of chunks of the same side length. Such a cube storage scheme is called the chunk-based MOLAP cube storage scheme. It gives data clustering effect so that all the dimensions are guaranteed to get a fair chance in terms of the query processing speed. In order to achieve high space utilization, sparse chunks are further compressed. Due to data compression, the relative position of chunks cannot be obtained in constant time without using indexes. In this paper, we propose a bitmap index for chunk-based MOLAP cubes. The index can be constructed along with the corresponding cube generation. The relative position of chunks is retained in the index so that chunk retrieval can be done in constant time. We placed in an index block as many chunks as possible so that the number of index searches is minimized for OLAP operations such as range queries. We showed the proposed index is efficient by comparing it with multidimensional indexes such as UB-tree and grid file in terms of time and space.

• Journal of Korea Game Society
• /
• v.18 no.5
• /
• pp.5-14
• /
• 2018

A lot of research has been done on image recognition technique for planar images and the performance has also been improved. However, it is difficult to recognize objects in spherical panoramic images or images in special form which are given in various environments because of the spherical distortion given in different form from the planar case. In this paper, we show that the neural network recognition approach can be used for object recognition in spherical image and suggest a method of using cubemap transform in order to increase recognition accuracy in spherical image.

• Journal of KIISE:Databases
• /
• v.33 no.4
• /
• pp.353-362
• /
• 2006

A data warehouse is a data repository that enables users to store large volume of data and to analyze it effectively. In this research, we investigate an algorithm to establish a multidimensional data cube which is a powerful analysis tool for the contents of data warehouses and databases. There exists an inevitable retrieval overhead in a multidimensional data cube due to the sparsity of the cube. In this paper, we propose a dense sub-cube extraction algorithm that identifies dense regions from a large sparse data cube and constructs the sub-cubes based on the dense regions found. It reduces the retrieval overhead remarkably by retrieving those small dense sub-cubes instead of scanning a large sparse cube. The algorithm utilizes the bitmap and histogram based techniques to extract dense sub-cubes from the data cube, and its effectiveness is demonstrated via an experiment.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2014.04a
• /
• pp.715-718
• /
• 2014

맵리듀스(MapReduce)는 대용량 데이터의 병렬 처리에 사용되는 프로그래밍 모델이다. 데이터 큐브(data cube)는 대용량 데이터의 다차원 분석에 널리 사용되는 연산자로서, 주어진 차원 애트리뷰트들의 모든 가능한 조합에 대한 group-by 를 계산한다. 차원 애트리뷰트가 n 개일 때, 데이터 큐브는 총 $2^n$ 개의 group-by 를 계산한다. 본 논문은 맵리듀스 환경에서 데이터 큐브를 효율적으로 계산하는 방법을 제안한다. 제안 방법은 $2^n$ 개의 group-by 를 분할하고 이들을 ${\lceil}n/2{\rceil}$개의 맵리듀스 잡(job)을 통해 단계적으로 계산한다. 제안 방법은 각 맵리듀스 잡에서 맵 함수가 출력하는 중간결과의 크기를 최소화함으로써 총 계산 비용을 크게 줄인다. 실험을 통해 제안 방법은 기존 방법에 비해 데이터 큐브를 더 빠르게 계산함을 보인다.

• Journal of Broadcast Engineering
• /
• v.22 no.3
• /
• pp.401-404
• /
• 2017

The 360 VR video has a format of a stereoscopic shape such as an isometric shape or a cubic shape or a cubic shape. Although these formats have different characteristics, they have in common that the resolution is higher than that of a normal 2D video. Therefore, it takes much longer time to perform coding/decoding on 360 VR video than 2D Video, so parallel processing techniques are essential when it comes to coding 360 VR video. HEVC, the state of art 2D video codec, uses Wavefront Parallel Processing (WPP) technology as a standard for parallelization. This technique is optimized for 2D videos and does not show optimal performance when used in 3D videos. Therefore, a suitable method for WPP is required for 3D video. In this paper, we propose WPP coding/decoding method which improves WPP performance on cube map format 3D video. The experiment was applied to the HEVC reference software HM 12.0. The experimental results show that there is no significant loss of PSNR compared with the existing WPP, and the coding complexity of 15% to 20% is further reduced. The proposed method is expected to be included in the future 3D VR video codecs.

• Journal of Korea Game Society
• /
• v.19 no.6
• /
• pp.49-60
• /
• 2019

Interior Mapping has been used to reduce graphics resources. In this paper, rendering speed(FPS), the number of polygons, shader complexity and each resource size of Interior Mapping were compared to those of actual modeling in order to examine the performance of 3D games when the technology is adapted by utilizing Unreal Engine 4. In addition, for the efficient application, the difference in performance according to the resolution and detail of cube map texture was verified.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2004.05a
• /
• pp.688-691
• /
• 2004

본 논문에서는 메쉬, 복셀, 골격 데이터를 포함하는 복합적인 옥트리 기반의 형상 표현을 이용하여 해마의 형상을 분석하기 위한 효과적인 방법을 제공한다. 먼저, 자기공명영상으로부터 분할된 해마 영역에 마칭큐브 알고리즘을 적용하여 다단계 메쉬 데이터를 생성한다. 이렇게 생성된 메쉬 모델을 하드웨어 깊이맵을 이용한 복셀화 과정을 통하여, 중간 단계의 이진 복셀 표현으로 변환한다. 마지막으로 광선 추적 방법에 의해 추출된 샘플 메쉬들에 대하여 L2 Norm을 계산함으로써 형상 특징을 생성한다. 본 연구에서 제시한 방법은 사용자 피킹 인터페이스를 이용하여 국부적 부위에서의 계층적 형상 분석을 가능하게 한다. 또한 계층적 Level-of-Detail 접근방법은 정확도를 유지하며 형상분석의 소요 시간을 절약하도록 한다.