• Journal of the Korean Society for Precision Engineering
• /
• v.23 no.10
• /
• pp.22-29
• /
• 2006

This paper considers the lens distortions such as a fisheye distortion and a perspective distortion. While a fisheye lens has a wide field-of-view, it causes a large distortion to the images. Regardless of a fisheye lens or a rectilinear lens, a lens generates perspective distortion in a vertical direction when the lens views in an upward direction or downward direction. These distortions deform images differently from human visual functions. Therefore, this paper presents a method to correct the distortions, and whereby, the research in this paper enlarges choices of images to image processing algorithm that may select the distorted images and the corrected images depending on applications. An infinite polynomial model is employed in the fisheye radial distortion correction, and the vertical perspective distortion correction is done by using a vanishing point. The methods introduced in this paper are implemented on the images captured by a rear-view camera installed on a vehicle and showed their robustness of the correction.

• Journal of the Optical Society of Korea
• /
• v.14 no.3
• /
• pp.245-259
• /
• 2010

We have developed mathematically precise image-processing algorithms for extracting panoramic images from fisheye images. Furthermore, we have successfully built a DSP-based panoramic camera employing single fisheye lens.

• Journal of the Optical Society of Korea
• /
• v.14 no.2
• /
• pp.109-120
• /
• 2010

We have developed mathematically precise image-processing algorithms for extracting rectilinear images from fisheye images as well as digital pan/tilt/zoom technology. Using this technology, vertical lines always appear as vertical lines in the panned and/or tilted images. Furthermore, polygonal panoramic images composed of multiple rectilinear images have been obtained using the developed digital pan/tilt technology.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.20-30
• /
• 2022

Recently, as VR (Virtual Reality) technology has been in the spotlight, 360° panoramic images that can view lively VR contents are attracting a lot of attention. Image stitching technology is a major technology for producing 360° panorama images, and many studies are being actively conducted. Typical stitching algorithms are based on feature point-based image stitching. However, conventional feature point-based image stitching methods have a problem that stitching results are intensely affected by feature points. To solve this problem, deep learning-based image stitching technologies have recently been studied, but there are still many problems when there are few overlapping areas between images or large parallax. In addition, there is a limit to complete supervised learning because labeled ground-truth panorama images cannot be obtained in a real environment. Therefore, we produced three fisheye images with different camera centers and corresponding ground truth image through carla simulator that is widely used in the autonomous driving field. We propose image stitching model that creates a 360° panorama image with the produced fisheye image. The final experimental results are virtual datasets configured similar to the actual environment, verifying stitching results that are strong against various environments and large parallax.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.10
• /
• pp.84-91
• /
• 2016

Most recently, the technology of around view monitoring(AVM) system or the security systems could provide users with images by using a fisheye lens. The filmed images through fisheye lens have an advantage of providing a wider range of scenes. On the other hand, filming through fisheye lens also has disadvantages of distorting images. Especially, it causes the sharpness of images to degrade because the edge of images is out of focus. The influence of a blur still remains at the end of the range when the super-resolution techniques is applied in order to enhance the sharpness. It degrades the clarity of high resolution images and occurs artifacts, which leads to deterioration in the performance of super-resolution algorithm. Therefore, in this paper we propose self-similarity-based pre-processing method to improve the sharpness at the edge. Additionally, we implement the acceleration in the GPU environment of entire algorithm and verify the acceleration.

• Journal of Conservation Science
• /
• v.37 no.4
• /
• pp.312-321
• /
• 2021

Underwater archaeology relies heavily on photography and video image recording during surveillances and excavations like ordinary archaeological studies on land. All underwater images suffer poor image quality and distortions due to poor visibility, low contrast and blur, caused by differences in refractive indices of water and air, properties of selected lenses and shapes of viewports. In the Yellow Sea (between mainland China and the Korean peninsula), the visibility underwater is far less than 1 m, typically in the range of 30 cm to 50 cm, on even a clear day, due to very high turbidity. For photographing 1 m x 1 m grids underwater, a very wide view angle (180°) fisheye lens with an 8 mm focal length is intentionally used despite unwanted severe barrel-shaped image distortion, even with a dome port camera housing. It is very difficult to map wide underwater archaeological excavation sites by combining severely distorted images. Development of practical compensation methods for distorted underwater images acquired through the fisheye lens is strongly desired. In this study, the source of image distortion in underwater photography is investigated. We have identified the source of image distortion as the mismatching, in optical axis and focal points, between dome port housing and fisheye lens. A practical image distortion compensation method, using customized image processing software, was explored and verified using archived underwater excavation images for effectiveness in underwater archaeological applications. To minimize unusable area due to severe distortion after distortion compensation, practical underwater photography guidelines are suggested.

• Proceedings of the IEEK Conference
• /
• 2008.06a
• /
• pp.963-964
• /
• 2008

Recently, omni-directional cameras are broadly used due to their wide field of view. Fisheye camera is one of them. This paper proposes the system which uses a fisheye camera for face recognition and analyzes its advantages. Since face images taken with a fisheye camera are affected by perspective distortion and radial distortion, we suggest a two-step method for removing those distortions from the face images.

• Journal of Korea Multimedia Society
• /
• v.22 no.2
• /
• pp.128-135
• /
• 2019

This paper presents a deep learning-based object detection method for recognizing vehicles in images acquired through cameras installed on ceiling of underground parking lot. First, we present an image enhancement method, which improves vehicle detection performance under dark lighting environment. Second, we present a new CNN-based multiscale classifiers for detecting vehicles in images acquired through cameras with fisheye lens. Experiments show that the presented vehicle detector has better performance than the conventional ones.

• Journal of Korea Multimedia Society
• /
• v.20 no.5
• /
• pp.769-781
• /
• 2017

Deep learning networks like Convolutional Neural Networks (CNNs) show successful performances in many computer vision applications such as image classification, object detection, and so on. For implementation of deep learning networks in embedded system with limited processing power and memory, deep learning network may need to be simplified. However, simplified deep learning network cannot learn every possible scene. One realistic strategy for embedded deep learning network is to construct a simplified deep learning network model optimized for the scene images of the installation place. Then, automatic training will be necessitated for commercialization. In this paper, as an intermediate step toward automatic training under fisheye camera environments, we study more precise human localization in fisheye images, and propose an accurate human localization method, Automatic Ground-Truth Labelling Method (AGTLM). AGTLM first localizes candidate human object bounding boxes by utilizing GoogLeNet-LSTM approach, and after reassurance process by GoogLeNet-based CNN network, finally refines them more correctly and precisely(tightly) by applying saliency object detection technique. The performance improvement of the proposed human localization method, AGTLM with respect to accuracy and tightness is shown through several experiments.

• IEIE Transactions on Smart Processing and Computing
• /
• v.2 no.6
• /
• pp.339-344
• /
• 2013

This paper presents an automatic focal length estimation method to correct the fisheye lens distortion in a spatially adaptive manner. The proposed method estimates the focal length of the fisheye lens by generating two reference focal lengths. The distorted fisheye lens image is finally corrected using the orthographic projection model. The experimental results showed that the proposed focal length estimation method is more accurate than existing methods in terms of the loss rate.