A5009320046- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Image Processing Techniques and Applications
- Advanced Image Fusion Techniques
- Advanced Neural Network Applications
- Remote Sensing and LiDAR Applications
- Domain Adaptation and Few-Shot Learning
- Chaos-based Image/Signal Encryption
- Advanced Vision and Imaging
- Advanced X-ray and CT Imaging
- Computer Graphics and Visualization Techniques
- Image Enhancement Techniques
- Image and Object Detection Techniques
- Advanced Image and Video Retrieval Techniques
Korea University
2019-2022
ETH Zurich
2019
This paper reviews the first AIM challenge on mapping camera RAW to RGB images with focus proposed solutions and results. The participating teams were solving a real-world photo enhancement problem, where goal was map original low-quality from Huawei P20 device same photos captured Canon 5D DSLR camera. considered problem embraced number of computer vision subtasks, such as image demosaicing, denoising, gamma correction, resolution sharpness enhancement, etc. target metric used in this...
Recent research on learning a mapping between raw Bayer images and RGB has progressed with the development of deep convolutional neural network.A challenging data set namely Zurich Raw-to-RGB (ZRR) been released in AIM 2019 raw-to-RGB challenge.In ZRR, input target are captured by two different cameras thus not perfectly aligned.Moreover, camera metadata such as white balance gains color correction matrix provided, which makes challenge more difficult.In this paper, we explore an effective...
Images captured from real-world environments often include blur artifacts resulting camera movement, dynamic object motion, or out-of-focus. Although such are inevitable, most detection methods do not have special considerations for them; therefore, they may fail to detect objects in blurry images. One possible solution is applying image deblurring prior detection. However, this computationally demanding and its performance heavily depends on results. In study, we propose a novel blur-aware...
X-ray baggage inspection has been widely used for maintaining airport and transportation security. Towards automated inspection, recent deep learning-based methods have attempted to detect hazardous objects directly from images. Since it is challenging collect a large number of training images real-world environments, most previous rely on image synthesis data generation. However, these randomly combine foreground background images, restricting the effectiveness synthetic object detection....
The single shot multi-box detector (SSD) is one of the first real-time detectors, which uses a convolutional neural network (CNN) and achieves state-of-the-art detection performance. However, owing to semantic gap between each feature layer CNN, SSD has room for improvement. In this paper, we propose novel training scheme enhance performance SSD. object detection, ground truth (GT) box bounding enclosing an boundary. To improve level map, generate additional GT boxes by zooming in out from...
Lossy video compression achieves coding gains at the expense of quality loss decoded images. Owing to success deep learning techniques, especially convolutional neural networks (CNNs), many artifacts reduction (CAR) techniques have been used significantly improve images by applying CNNs which are trained predict original artifact-free from Most existing standards control ratio using a quantization parameter (QP), so is strongly QP-dependent. Training individual for predetermined QPs one...
본 논문에서는 기존의 MLCA(Maximum Length CA) 및 여원 MLCA 기반의 난수열을 이용한 영상 암호화 방법의 문제점을 제시하고 이를 해결하기 위한 방법을 제안한다. 방법은 영상에서 인접한 픽셀간의 유사한 색상 값을 가지는 특성으로 인해 암호화된 영상에 원영상의 윤곽이 나타나는 문제점이 발생한다. 생성하고, 이용해서 픽셀의 값뿐만 아니라 공간좌표를 무질서하게 변환함으로써 영상을 암호화한다. 이러한 데이터를 시각화하기 영상의 기본 정보인 값과 변환시키기 때문에 기존 방법보다 향상된 결과를 얻을 수 있다. 히스토그램 키 공간 분석을 통해 안정성을 평가한다. This paper presents a problem of existing encryption methods using pseudo-random numbers based on or complemented and proposes method to resolve this problem. The have which...
Image denoising is a classical and essential task in consumer electronics equipped with cameras. Recently, the convolutional neural network (CNN)-based methods have been widely studied. These adopt single-scale features to separate image structures from noisy observation. Single-scale features, however, limitation covering full characteristics of at different scales. In this paper, we propose novel that makes use multi-scale feature pyramid where each map represents structure We then combine...