• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.01a
• /
• pp.257-260
• /
• 2021

본 논문에서는 카메라의 포커싱과 아웃포커싱에 의해 이미지에서 뿌옇게 표현되는 DoF(Depth of field, 피사계 심도) 영역을 쿼드트리(Quadtree) 기반의 합성곱 신경망을 통해 빠르게 찾는 방법을 제안한다. 우리의 접근 방식은 RGB채널기반의 상호-상관 필터를 이용하여 DoF영역을 이미지로부터 효율적으로 분류하고, 적응형 트리인 쿼드트리를 기반으로 유의미한 영역만을 분류한다. 이 과정에서 손실 없이 온전하게 DoF영역을 추출하기 위한 필터링 과정을 거친다. 이러한 과정에서 얻어진 이미지 패치들은 전체 이미지에 비해 적은 영역으로 나타나며, 이 적은 개수의 패치들을 이용하여 네트워크 단계에서 사용할 이미지-DoF가중치 맵 데이터 쌍을 설정한다. 네트워크 과정에서 학습할 때 사용되는 데이터는 이미지와 상호-상관 필터 기반으로 추출된 DoF 가중치 맵을 이용한다. 본 논문에서 제안하는 쿼드트리 기반 합성곱 신경망은 이미지로부터 포커싱과 아웃포커싱된 DoF영역을 자동으로 추출하는 과정을 학습시키기 위해 사용된다. 결과적으로 학습에 필요한 데이터 영역이 줄어듦으로써 학습 시간과 메모리를 절약했으며, 테스트 결과로 얻은 DoF 가중치 이미지는 입력 이미지에서 DoF영역을 더욱더 빠른 시간 내에 찾아낸다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2020.07a
• /
• pp.429-432
• /
• 2020

본 논문에서는 카메라의 포커싱과 아웃포커싱에 의해 이미지에서 뿌옇게 표현되는 DoF(Depth of field, 피사계 심도) 영역을 합성곱 신경망을 통해 찾는 방법을 제안한다. 우리의 접근 방식은 RGB채널기반의 상호-상관 필터를 이용하여 DoF영역을 이미지로부터 효율적으로 분류하고, 합성곱 신경망 네트워크에 학습하기 위한 데이터를 구축하며, 이렇게 얻어진 데이터를 이용하여 이미지-DoF가중치 맵 데이터 쌍을 설정한다. 학습할 때 사용되는 데이터는 이미지와 상호-상관 필터 기반으로 추출된 DoF 가중치 맵을 이용하며, 네트워크 학습 단계에서 수렴률을 높이기 위해 스무딩을 과정을 한번 더 적용한 결과를 사용한다. 본 논문에서 제안하는 합성곱 신경망은 이미지로부터 포커싱과 아웃포커싱된 DoF영역을 자동으로 추출하는 과정을 학습시키기 위해 사용된다. 테스트 결과로 얻은 DoF 가중치 이미지는 입력 이미지에서 DoF영역을 빠른 시간 내에 찾아내며, 제안하는 방법은 DoF영역을 사용자의 ROI(Region of interest)로 활용하여 NPR렌더링, 객체 검출 등 다양한 곳에 활용이 가능하다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2010.07a
• /
• pp.435-437
• /
• 2010

렌즈의 피사계심도 (Depth of Field)가 낮은 카메라로 영상을 획득 했을 때, 한 영상 내에도 in-focus 영역과 out of-focus 영역이 동시에 존재하게 된다. 따라서 영상을 복원하기 위해 in-focus 영역과 out-of-focus 영역을 구분하는 focus measure가 필요하게 된다. 기존 focus measure 알고리즘은 획득된 영상의 intensity 값의 절대적인 변화나 고주파수 성분 값에 따라 in-focus와 out-of-focus를 결정하기 때문에 out-of-focus 영역이 smooth 하지 않을 경우에는 in-focus 영역이라 잘못 판단할 수 있을 뿐만 아니라 잡음에 민감한 단점을 가진다. 본 논문에서는 기존 알고리즘의 한계점을 보완하는 연구 방향을 제시한다.

• Journal of Broadcast Engineering
• /
• v.16 no.5
• /
• pp.883-893
• /
• 2011

In this paper, we propose a novel focus measure that determines in-focus and out-of-focus region in an image. In addition, we achieved extended depth of field by blending the acquired image and Wiener filtered image using a decision map based on the designed focus measure. Since conventional focus measures are based on the amount of high frequency components in an acquired image, the measures may not be accurate if there exist high frequency components in out-of-focused region. To overcome the problem, we designed the novel focus measure based on effective band pass filtering. In simulations and experiments, the proposed method showed better performance than existing methods.

• Proceedings of the Optical Society of Korea Conference
• /
• 2007.02a
• /
• pp.72-73
• /
• 2007

• Korean Journal of Optics and Photonics
• /
• v.33 no.6
• /
• pp.295-302
• /
• 2022

In this paper, a wedge prism application method was studied to design a full-high-definition (FHD)-class high-resolution flexible endoscope. In the case of the conventional flexible endoscope optical system, the F number is made large or a liquid lens is applied to obtain the same imaging performance in a wide depth of field. However, there is a problem in that the diameter of the optical system increases because an additional light guide and equipment are required. To solve this problem, two wedge prisms were applied to the flexible endoscope optical system to adjust the image distance for each object distance. First, two wedge prisms were symmetrically placed on the designed endoscopic optical system. An image distance satisfying the target imaging performance according to each objective distance was derived. Next, the wedge prism decenter value for controlling the image distance was derived. By combining these two data, a wedge prism decenter value that satisfied the target imaging performance at each object distance was applied in multi configurations. As a result of the optimal design applied with the wedge prism, a target imaging performance of more than 20% of the modulation transfer function for a resolution of 178 cycles/mm was satisfied in the entire depth of field of 100 mm-7 mm.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.7
• /
• pp.169-179
• /
• 2014

In this paper, we propose a DOF (Depth of Field) correction technique by determining the values of the internal parameters of a 3-D camera which consists of stereoscopic cameras of different optical properties. If there is any difference in the size or the depth range of focused objects in the left and right stereoscopic images, it could cause visual fatigue to human viewers. The object size of in the stereoscopic image is corrected by the LUT of zoom lenses, and the forward and backward DOF are corrected by the object distance. Then the F-numbers are determined to adjust the optical properties of the camera for DOF correction. By applying the proposed technique to a main-sub type 3-D camera using a GUI-based DOF simulator, the DOF of the camera is automatically corrected.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.3
• /
• pp.51-57
• /
• 2021

In this paper, we propose a method to find the DoF(Depth of field) that is blurred in an image by focusing and out-focusing the camera through a efficient convolutional neural network. Our approach uses the RGB channel-based cross-correlation filter to efficiently classify the DoF region from the image and build data for learning in the convolutional neural network. A data pair of the training data is established between the image and the DoF weighted map. Data used for learning uses DoF weight maps extracted by cross-correlation filters, and uses the result of applying the smoothing process to increase the convergence rate in the network learning stage. The DoF weighted image obtained as the test result stably finds the DoF region in the input image. As a result, the proposed method can be used in various places such as NPR(Non-photorealistic rendering) rendering and object detection by using the DoF area as the user's ROI(Region of interest).

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.10
• /
• pp.63-71
• /
• 2009

In this study, we propose a revised dynamic ROI (Region-of-Interest) coding method in which the focused ROI is automatically extracted without help from users during the recovery process of low DOF (Depth-of-Field) JPEG2000 image. The proposed method creates edge mask information using high frequency sub-band data on a specific level in DWT (Discrete Wavelet Transform), and then identifies the edge code block for a high-speed ROI extraction. The algorithm scans the edge mask data in four directions by the unit of code block and identifies the edge code block simply and fastly using a edge threshold. As the results of experimentation applying for Implicit method, the proposed method showed the superiority in the side of speed and quality comparing to the existing methods.

• Journal of Digital Contents Society
• /
• v.9 no.2
• /
• pp.187-192
• /
• 2008

This study carried out researches on the techniques which are applicable to express sense of depth in the process of producing lofty image contents, and it's verification. Techniques for expression of sense of depth are how to use highlight and shadow, how to use depth of field and lenses, and how to use camera motion, zoom and dolly. For the practical application in all cases, I examine the preference frequency, and most of the pictures that techniques are adapted are selected. As a result we can verify the way of using lenses is the most convenient way in the expression of 'sense of depth'.