Bo-Hao Chen

Visibility Restoration of Single Hazy Images Captured in Real-World Weather Conditions

The visibility of outdoor images captured in inclement weather is often degraded due to the presence of haze, fog, sandstorms, and so on. Poor visibility caused by atmospheric phenomena in turn causes failure in computer vision applications, such as outdoor object recognition systems, obstacle detection systems, video surveillance systems, and intelligent transportation systems. In order to solve this problem, visibility restoration (VR) techniques have been developed and play an important role in many computer vision applications that operate in various weather conditions. However, removing haze from a single image with a complex structure and color distortion is a difficult task for VR techniques. This paper proposes a novel VR method that uses a combination of three major modules: 1) a depth estimation (DE) module; 2) a color analysis (CA) module; and 3) a VR module. The proposed DE module takes advantage of the median filter technique and adopts our adaptive gamma correction technique. By doing so, halo effects can be avoided in images with complex structures, and effective transmission map estimation can be achieved. The proposed CA module is based on the gray world assumption and analyzes the color characteristics of the input hazy image. Subsequently, the VR module uses the adjusted transmission map and the color-correlated information to repair the color distortion in variable scenes captured during inclement weather conditions. The experimental results demonstrate that our proposed method provides superior haze removal in comparison with the previous state-of-the-art method through qualitative and quantitative evaluations of different scenes captured during various weather conditions.

An Efficient Visibility Enhancement Algorithm for Road Scenes Captured by Intelligent Transportation Systems

The visibility of images of outdoor road scenes will generally become degraded when captured during inclement weather conditions. Drivers often turn on the headlights of their vehicles and streetlights are often activated, resulting in localized light sources in images capturing road scenes in these conditions. Additionally, sandstorms are also weather events that are commonly encountered when driving in some regions. In sandstorms, atmospheric sand has a propensity to irregularly absorb specific portions of a spectrum, thereby causing color-shift problems in the captured image. Traditional state-of-the-art restoration techniques are unable to effectively cope with these hazy road images that feature localized light sources or color-shift problems. In response, we present a novel and effective haze removal approach to remedy problems caused by localized light sources and color shifts, which thereby achieves superior restoration results for single hazy images. The performance of the proposed method has been proven through quantitative and qualitative evaluations. Experimental results demonstrate that the proposed haze removal technique can more effectively recover scene radiance while demanding fewer computational costs than traditional state-of-the-art haze removal techniques.

A Novel Visibility Restoration Algorithm for Single Hazy Images

The visibility of outdoor images captured in inclement weather will become degraded due to the presence of haze, fog, mist, and so on. Poor visibility caused by atmospheric phenomenon in turn causes failure in computer vision applications, such as outdoor object recognition systems, obstacle detection systems, video surveillance systems, and intelligent transportation systems. In order to solve this problem, visibility restoration techniques have been developed and play an important role in many computer vision applications. However, complete haze removal from a single image with a complex structure is difficult for visibility restoration techniques to achieve. This paper proposes a novel visibility restoration method which utilizes a combination of the median filter operation and the dark channel prior in order to achieve effective haze removal in a single image with a complex structure. The experimental results demonstrate that our proposed method provides superior haze removal in comparison to the previous state-of-the-art method through visual evaluation of different scenes.

Visibility Enhancement of Single Hazy Images Using Hybrid Dark Channel Prior

Outdoor images captured during inclement weather conditions generally exhibit visibility degradation. Localized light sources often result from activation of streetlights and vehicle headlights and are common scenarios in these conditions. The presence of localized light sources in hazy images may cause the generation of over saturation artifacts when those images are restored by traditional state-of-the-art haze removal techniques. Therefore, we propose a novel haze removal approach based on the proposed hybrid dark channel prior technique in order to remedy the problems associated with localized light sources during image restoration. The overall results show that the proposed haze removal approach can recover haze-free images more effectively than can the other previous state-of-the-art haze removal approach while avoiding over-saturation.

Improved global motion estimation via motion vector clustering for video stabilization

Video stabilization technique is often used in handheld multimedia devices, whereas the difficulties in the accurate extraction aspect of global motion vectors restrict its development. This paper proposes a novel video stabilization approach that is based on the shortest spanning path clustering algorithm for effective and reliable estimation of the global motion vectors. As demonstrated in our experimental results, the proposed approach achieves superior stabilized effectiveness compared with the other state-of-the-art approaches based on both qualitative and quantitative measurements.

Error-Optimized Sparse Representation for Single Image Rain Removal

Feature extraction and visual attention modeling of captured images are often used in outdoor imaging systems; however, corruption of images by rain streaks poses difficulties that restrict the development of these techniques. In this paper, we propose a novel rain streak removal method that is based on an error-optimized sparse representation (EOSR) model developed in this study. Derived from the sparse representation model, the proposed EOSR model can be used to compute each image patch by considering the dynamic patch error constraints, which can then be optimized using nondominated sorting-based genetic algorithms through the multiobjective pursuit of single-image rain streak removal. In contrast to previously used methods that focus on dictionary partition for rain streak removal, the proposed model flexibly represents each image patch on the basis of optimized patch error constraints. Experimental results derived through qualitative and quantitative evaluations indicated that the proposed model could efficiently remove rain streaks from each image patch; thus, facilitating the reconstruction of a visually superior rain-free image compared with those produced by other state-of-the-art methods.

An automatic motion detection algorithm for transport monitoring systems

Objects need to be analyzed in the video surveillance system, while motion detection can be applied to define the analyzable area. This paper proposes a novel motion detection algorithm with background model generation. In order to accurately generate background model, 4-connectivity function is directly used to approximately label the background region. The labeling function makes the background model self-adaptive as object pixels can be roughly ignored for updating. Similarly, this function is also applied to approximately select the foreground region after background subtraction. Finally, objects can be detected by direct threshold function from the labeled foreground region. For measuring the quantitative accuracy, Similarity and F1 are used as two accuracy metrics. As a result, the proposed method produces a substantial degree of efficacy higher than those produced by other state-of-the-art methods.

Haze Removal Using Radial Basis Function Networks for Visibility Restoration Applications

Restoration of visibility in hazy images is the first relevant step of information analysis in many outdoor computer vision applications. To this aim, the restored image must feature clear visibility with sufficient brightness and visible edges, while avoiding the production of noticeable artifacts. In this paper, we propose a haze removal approach based on the radial basis function (RBF) through artificial neural networks dedicated to effectively removing haze formation while retaining not only the visible edges but also the brightness of restored images. Unlike traditional haze-removal methods that consist of single atmospheric veils, the multiatmospheric veil is generated and then dynamically learned by the neurons of the proposed RBF networks according to the scene complexity. Through this process, more visible edges are retained in the restored images. Subsequently, the activation function during the testing process is employed to represent the brightness of the restored image. We compare the proposed method with the other state-of-the-art haze-removal methods and report experimental results in terms of qualitative and quantitative evaluations for benchmark color images captured in typical hazy weather conditions. The experimental results demonstrate that the proposed method is able to produce brighter and more vivid haze-free images with more visible edges than can the other state-of-the-art methods.

Parametric Representation of Objects in Color Space Using One-Class Classifiers

Two new approaches to parametrization of specific (flame representative) part of a color space, labeled by an expert, are presented. The first concept is to apply D. Tax’s one-class classifier as a steerable descriptor of such a complex volumetric structure. The second concept is based on approximation of the training data by a set of elliptic cylinders arranged along the principal components. Parameters of such elliptic cylinders describe the training set. The efficiency of the approaches has been proven by experimental study which let allowed us to compare the standard Gaussian Mixture Model based approach with the two proposed in the paper.