안개량 오추정 영역 보정을 이용한 개선된 Dark Channel Prior 안개 제거 알고리즘

김종현(Jong-Hyun Kim),차형태(Hyung-Tai Cha) 한국방송·미디어공학회 2016 방송공학회논문지 Vol.21 No.5

As a result of reducing color information and edge information, object distinction in haze image occurs with difficulty. One of the famous defogging algorithm is haze removal by using ‘Dark Channel Prior(DCP)’, which is used to predict for transmission rate using color information of an image and eliminates haze from the image. But, In case that haze rate is estimated under color information, there is a miscalculated issue which is posed by haze rate and transmission in area with high brightness such as a white object or a light source. In this paper, We deal with a miscalculated issue by correcting from around haze rate, after application of color normalization used by main white part of image haze. Moreover, We calculation improved transmission based on the result of improved haze rate estimation. And then haze image quality is developed through refining transmission.

An Improved Haze Removal Algorithm Based on Genetic Fuzzy Clustering

Xiaoguang Li,Huiying Huang 보안공학연구지원센터 2015 International Journal of Hybrid Information Techno Vol.8 No.8

Aiming at the degeneration phenomenon of images taken in mist, according to the features of the degraded images, an improved haze removal algorithm based on genetic fuzzy clustering is presented in this paper after analyzing its defects and shortcomings. Firstly, the improved atmospheric scattering model is established. Secondly, the original image is converted into a standard image through the improved model, and then we present a new multi-scale image edge detection by genetic fuzzy clustering, Based on this observation, we can use the multi-scale image edge detection to estimate the haze thickness directly and recover a high quality haze-free image. The new algorithm uses good global search ability of the genetic algorithm, which will implement the transfer from the scene defogging problem into the optimal estimation problem under global contrast optimal point. Compared with other algorithms for degraded images, the improved haze removal algorithm not only detects image edge precisely, but also has better performance in situations of dense haze. Theoretical analysis and experimental results demonstrate that, the new algorithm improved in this paper are effective for removal of fog-degraded images, and can be applied to the practical situations.

A 4K-Capable Hardware Accelerator of Haze Removal Algorithm using Haze-relevant Features

이승민,강봉순 한국정보통신학회 2022 Journal of information and communication convergen Vol.20 No.3

The performance of vision-based intelligent systems, such as self-driving cars and unmanned aerial vehicles, is subject to weather conditions, notably the frequently encountered haze or fog. As a result, studies on haze removal have garnered increasing interest from academia and industry. This paper hereby presents a 4K-capable hardware implementation of an efficient haze removal algorithm with the following two improvements. First, the depth-dependent haze distribution is predicted using a linear model of four haze-relevant features, where the model parameters are obtained through maximum likelihood estimates. Second, the approximated quad-decomposition method is adopted to estimate the atmospheric light. Extensive experimental results then follow to verify the efficacy of the proposed algorithm against well-known benchmark methods. For real-time processing, this paper also presents a pipelined architecture comprised of customized macros, such as split multipliers, parallel dividers, and serial dividers. The implementation results demonstrated that the proposed hardware design can handle DCI 4K videos at 30.8 frames per second.

Dual DCP 및 적응적 밝기 보정을 통한 단일 영상 기반 안개 제거 알고리즘

김종호(Jongho Kim) 한국산학기술학회 2018 한국산학기술학회논문지 Vol.19 No.11

본 논문에서는 효과적이고 저 복잡도를 갖는 단일 영상 기반의 안개 제거를 위하여 dual dark channel prior (DCP)와 적응적인 밝기 보정 기법을 이용하는 알고리즘을 제안한다. 작은 크기의 패치에 의한 dark channel은 영상의 에지 정보를 잘 보존하지만 국부적인 잡음 및 밝기 변화에 민감한 반면, 큰 크기의 패치에 의한 dark channel은 정확한 안개값을 추정하는데 유리하지만 블록 현상과 이로 인한 후광 효과는 안개 제거 성능을 떨어뜨린다. 이러한 문제를 해결하기 위하여 기존의 방법에서는 계산량 및 메모리 요구량이 큰 matting 기법을 활용한 반면, 제안하는 방법은 크기가 다른 패치로부터 구한 dark channel을 합성하여 dual DCP를 구성하고, 이를 이용하여 안개를 제거함으로써 적은 계산량 및 메모리 요구량을 달성한다. 또한 안개 성분을 제거한 영상에 적응적 밝기 보정 기법을 적용하여 영상에 포함된 객체가 선명하게 보존될 수 있도록 한다. 안개 성분이 포함된 다양한 영상에 대해 수행한 실험 결과 제안하는 안개 제거 기법이 기존의 방법에 비해 안개 제거 성능이 우수하면서 계산량과 메모리 요구량이 감소함을 알 수 있다. This paper proposes an effective single-image haze-removal algorithm with low complexity by using a dual dark channel prior (DCP) and an adaptive brightness correction technique. The dark channel of a small patch preserves the edge information of the image, but is sensitive to noise and local brightness variations. On the other hand, the dark channel of a large patch is advantageous in estimation of the exact haze value, but halo effects from block effects deteriorate haze-removal performance. In order to solve this problem, the proposed algorithm builds a dual DCP as a combination of dark channels from patches with different sizes, and this meets low-memory and low-complexity requirements, while the conventional method uses a matting technique, which requires a large amount of memory and heavy computations. Moreover, an adaptive brightness correction technique that is applied to the recovered image preserves the objects in the image more clearly. Experimental results for various hazy images demonstrate that the proposed algorithm removes haze effectively, while requiring much fewer computations and less memory than conventional methods.

Improving Performance of Machine Learning-based Haze Removal Algorithms with Enhanced Training Database

Dat Ngo,강봉순 한국전기전자학회 2018 전기전자학회논문지 Vol.22 No.4

Haze removal is an object of scientific desire due to its various practical applications. Existing algorithms arefounded upon histogram equalization, contrast maximization, or the growing trend of applying machine learning inimage processing. Since machine learning-based algorithms solve problems based on the data, they usuallyperform better than those based on traditional image processing/computer vision techniques. However, to achievesuch a high performance, one of the requisites is a large and reliable training database, which seems to beunattainable owing to the complexity of real hazy and haze-free images acquisition. As a result, researchers arecurrently using the synthetic database, obtained by introducing the synthetic haze drawn from the standarduniform distribution into the clear images. In this paper, we propose the enhanced equidistribution, improving uponour previous study on equidistribution, and use it to make a new database for training machine learning-basedhaze removal algorithms. A large number of experiments verify the effectiveness of our proposed methodology

Improving Performance of Machine Learning-based Haze Removal Algorithms with Enhanced Training Database

Ngo, Dat,Kang, Bongsoon Institute of Korean Electrical and Electronics Eng 2018 전기전자학회논문지 Vol.22 No.4

Haze removal is an object of scientific desire due to its various practical applications. Existing algorithms are founded upon histogram equalization, contrast maximization, or the growing trend of applying machine learning in image processing. Since machine learning-based algorithms solve problems based on the data, they usually perform better than those based on traditional image processing/computer vision techniques. However, to achieve such a high performance, one of the requisites is a large and reliable training database, which seems to be unattainable owing to the complexity of real hazy and haze-free images acquisition. As a result, researchers are currently using the synthetic database, obtained by introducing the synthetic haze drawn from the standard uniform distribution into the clear images. In this paper, we propose the enhanced equidistribution, improving upon our previous study on equidistribution, and use it to make a new database for training machine learning-based haze removal algorithms. A large number of experiments verify the effectiveness of our proposed methodology.

영상의 화질을 실시간으로 개선하기 위한 안개 제거 하드웨어의 설계

이재동,김은경,김영형,이용환 한국정보기술학회 2014 한국정보기술학회논문지 Vol. No.

Recently, image processing systems are used in various fields. However, the low-grade quality of image caused by haze is a critical issue for image system to be solved. In this paper, we design hardware to remove haze in image in real time. We use an algorithm using dark channel prior that extracts the value of dark channel from image and calculates the transmission rate. After that, bilateral filter refines the image to increase the effect of the haze removal. We used the pipeline and parallel processing structure for real-time processing that increase the speed of calculation more than 10 times. Fixed point arithmetic is used to accelerate the computations and reduce the number of gates. The proposed architecture was implemented using FPGA to verify the function and measure the performance. 영상 시스템은 최근 다양한 분야에서 활용되고 있다. 그러나 안개 등에 의해 발생하는 화질 저하는 시스템의 제 기능 발휘를 위하여 반드시 해결되어야 할 문제이다. 본 논문에서는 영상 내에 존재하는 안개를 실시간으로 제거하기 위한 하드웨어를 설계한다. Dark channel prior를 이용한 알고리즘을 사용하여 안개가 존재하는 영상에서 dark channel 값을 추출하여 전달량을 계산하고 그 후 양방향 필터로 정화하여 안개 제거의 효과를 높인다. 실시간 처리를 위해 하드웨어의 병렬처리 구조와 파이프라인을 사용하여 기존의 연산의 속도를 10배 이상 증가시킨다. 또한 하드웨어 설계에 적합하도록 고정소수점 연산을 사용하여 게이트 크기를 줄이고 연산속도를 증가시킨다. 설계된 하드웨어는 FPGA로 구현하여 그 기능을 검증하고 성능을 측정하였다.

Sharpness-aware Evaluation Methodology for Haze-removal Processing in Automotive Systems

Seokha Hwang,Youngjoo Lee 대한전자공학회 2016 IEIE Transactions on Smart Processing & Computing Vol.5 No.6

This paper presents a new comparison method for haze-removal algorithms in nextgeneration automotive systems. Compared to previous peak signal-to-noise ratio–based comparisons, which measure similarity, the proposed modulation transfer function–based method checks sharpness to select a more suitable haze-removal algorithm for lane detection. Among the practical filtering schemes used for a haze-removal algorithm, experimental results show that Gaussian filtering effectively preserves the sharpness of road images, enhancing lane detection accuracy.

안개 영상의 블럭 결함 제거와 변위 맵을 이용한 평가

권오설(Oh-Seol Kwon) 한국방송·미디어공학회 2014 방송공학회논문지 Vol.19 No.5

In the case of a haze image, transferring the information of the original image is difficult as the contrast leans toward bright regions. Thus, dehazing algorithms have become an important area of study. Normally, since it is hard to obtain a haze-free image, the output image is qualitatively analyzed to test the performance of an algorithm. However, this paper proposes a quantitative error comparison based on reproducing the haze image using a disparity map. In addition, a Hidden Random Markov Model and EM algorithm are used to remove any block artifacts. The performance of the proposed algorithm is confirmed using a variety of synthetic and natural images.

영상의 화질을 실시간으로 개선하기 위한 안개 제거 하드웨어의 설계

이재동(Jae-Dong Lee),김은경(Eun-Kyoung Kim),김영형(Young-Hyung Kim),이용환(Yong-Hwan Lee) 한국정보기술학회 2014 한국정보기술학회논문지 Vol.12 No.2

Recently, image processing systems are used in various fields. However, the low-grade quality of image caused by haze is a critical issue for image system to be solved. In this paper, we design hardware to remove haze in image in real time. We use an algorithm using dark channel prior that extracts the value of dark channel from image and calculates the transmission rate. After that, bilateral filter refines the image to increase the effect of the haze removal. We used the pipeline and parallel processing structure for real-time processing that increase the speed of calculation more than 10 times. Fixed point arithmetic is used to accelerate the computations and reduce the number of gates. The proposed architecture was implemented using FPGA to verify the function and measure the performance.