• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2011.07a
• /
• pp.226-229
• /
• 2011

얼굴검출기술의 발전으로 인하여, 다양한 분야에 얼굴 검출을 활용한 기술이 이용되고 있다. 최근 Viola 와 Jones 의 얼굴검출 방법이 신뢰도 있는 검출률과 빠른 연산속도로 인하여 주로 이용되고 있다. 하지만 고해상도 이미지와 제한된 하드웨어를 사용하는 시스템의 경우, 실시간 처리가 어려워지는 문제가 있다. 본 논문에서는 이와 같은 문제를 해결하고자 거리 정보를 이용한 빠른 얼굴검출방법을 제안한다. 속도 개선을 위해 먼저 거리 정보를 이용하여 영상의 불필요한 부분을 제거하고, 피부색상정보를 이용하여 관심영역을 설정한다. 또 크기에 대응하기 위해 피라미드 이미지를 이용하는 방법 대신, 거리 정보를 이용하여 얼굴의 크기를 추정한다. 마지막으로 검색창 내의 거리 분산을 계산하여, 평평하거나 굴곡이 심한 영역을 제거함으로 얼굴 검출 속도를 개선하였다. 실험결과 기존 방법에 비해 더 빠른 검출속도와 유사한 검출성능을 확인할 수 있었다.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.4
• /
• pp.158-167
• /
• 2007

Indirect illumination plays an important role for realistic image synthesis. We present a novel realtime indirect illumination rendering technique using image pyramids. Hundreds of thousands of indirect point light sources are stored into images, and then they hierarchically clustered into quad-tree image pyramids. We also introduce a GPU based top-down and breadth-first traversal of the quad-trees to approximate the illumination with clusters (set of lights). All steps entirely run on the GPU in real-time. Result images demonstrate that our method represents diffuse interreflection, especially a color bleeding effect well. We achieved interactive frame rates of tens to hundreads, without any preprocessing. We can avoid artifacts caused by sampling, and our method is seven times faster than a recently proposed sampling based method.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.1
• /
• pp.50-56
• /
• 2020

With the rapid development of deep neural network theory and application research, the effect of face detection has been improved. However, due to the complexity of deep neural network calculation and the high complexity of the detection environment, how to detect face quickly and accurately becomes the main problem. This paper is based on the relatively simple model of the MTCNN model, using FDDB (Face Detection Dataset and Benchmark Homepage), LFW (Field Label Face) and FaceScrub public datasets as training samples. At the same time of sorting out and introducing MTCNN(Multi-Task Cascaded Convolutional Neural Network) model, it explores how to improve training speed and Increase performance at the same time. In this paper, the dynamic image pyramid technology is used to replace the traditional image pyramid technology to segment samples, and OHEM (the online hard example mine) function in MTCNN model is deleted in training, so as to improve the training speed.

• Journal of the Korean Vacuum Society
• /
• v.17 no.3
• /
• pp.226-233
• /
• 2008

The nucleation and evolution process of Ge nano-islands on Si(001) surfaces grown by chemical vapor deposition have been explored using atomic force microscopy (AFM). The Ge nano-islands are grown by exposing the substrates to a mixture of gasses GeH4 and H2 at pressure of 0.1-0.5Torr and temperatures of $600-650^{\circ}C$. The effect of growth conditions such as temperature, Ge thickness, annealing time on the shape, size, number density, and surface distribution was investigated. For Ge deposition greater than ${\sim}5$ monolayer (ML) with a growth rate of ${\sim}0.1ML/sec$ at $600^{\circ}C$, we observed island nucleation on the surface indicating the transition from strained layer to island structure. Further deposition of Ge led to shape transition from initial pyramid and hut to dome and superdome structure. The lateral average size of the islands increased from ${\sim}20nm$ to ${\sim}310nm$ while the number density decreased from $4{\times}10^{18}$ to $5{\times}10^8cm^{-2}$ during the shape transition process. In contrast, for the samples grown at a relatively higher temperature of $650^{\circ}C$ the morphology of the islands showed that the dome shape is dominant over the pyramid shape. The further deposition of Ge led to transition from the dome to the superdome shape. The evolution of shape, size, and surface distribution is related to energy minimization of the islands and surface diffusion of Ge adatoms. In particular, we found that the initially nucleated islands did not grow through long-range interaction between whole islands on the surface but via local interaction between the neighbor islands by investigation of the inter-islands distance.