• Publications
Show publication details

Farhadifard, Fahimeh; Zhou, Zhiliang; Lukas, Uwe von

Learning-based Underwater Image Enhancement with Adaptive Color Mapping

2015

ISPA 2015

International Symposium on Image and Signal Processing and Analysis (ISPA) <9, 2015, Zagreb, Croatia>

Blurring and color cast are two of the most challenging problems for underwater imaging. The poor quality hinders the automatic segmentation or analysis of images. In this paper, we describe an image enhancement method to reduce the blurring and color cast of the underwater medium. It is a twofolded approach; First, a color correction algorithm is applied to correct the color cast and produce a natural appearance of the sub-sea images. Second, a pair of learned dictionaries based on sparse representation are applied to sharpen the image and enhance the details. Our strategy is a single image approach that does not require additional knowledge of environment such as depth, distance object/camera or water quality. The experimental results show that the proposed method can efficiently enhance almost every underwater image; And offers a quality that is typically sufficient for the high level computer vision algorithms.

Show publication details

Gennies, Maria; Urban, Bodo [Gutachter]; Gutzeit, Enrico [Betreuer]; Zhou, Zhiliang [Betreuer]

Echtzeit-Gesichtsdetektion und -verschleierung in Videos unter praktischen Bedingungen

2014

Magdeburg, Univ., Diplomarbeit, 2014

Die Detektion und Verschleierung von Gesichtern in Videoströmen verschiedener Kameras aus Innen- und Außengeländebereichen stellt eine besondere Herausforderung dar. In der vorliegenden Arbeit werden die theoretischen Grundlagen dieser Fragestellung näher betrachtet, ein umfassendes Konzept zur automatischen Gesichtsdetektion und -verschleierung unter den gegebenen praktischen Bedingungen entwickelt sowie die Umsetzung und Resultate der grundlegenden Systemarchitektur erläutert. Das vorgestellte prototypische System besteht aus einer Kombination von Bildvorverarbeitungsmethoden, der fensterbasierten Merkmalsextraktion mit Haar-like Features, der Lernkomponente AdaBoost, einem farbbasierten Tracking mit dem CAMShift-Algorithmus sowie einer Verschleierung. Weiterhin werden Alternativmethoden, Erweiterungen und Optimierungsmöglichkeiten zur Steigerung der Robustheit und Echtzeitfähigkeit des Systems vorgestellt.

Show publication details

Henke, Birgit; Zhou, Zhiliang [Betreuer]

Farbkorrektur-Algorithmen für die Unterwasserphotographie und Marktanalyse für eine Kameraerweiterung

2013

Rostock, Univ., Bachelor Thesis, 2013

Die Unterwasserphotographie leidet unter einer starken Farbverfälschung durch unterschiedliche, frequenzabhängige Lichtdämpfungen. Für die nachträgliche Bearbeitung der Bilder reicht die Leistung der bestehenden Farbkorrekturalgorithmen nicht aus und Fehlapproximationen, die zu Instabilität und Bildfehlern führen, sind die Folge. Aus diesem Grund wurde im Rahmen dieser Arbeit ein Lösungsansatz entwickelt, der verschiedene Farbkorrekturalgorithmen durch einen Multilevel-Fusion-Prozess kombiniert. Das Ziel stellt dabei die Nutzung der verschiedenen Stärken der einzelnen Algorithmen dar, die in ein gemeinsames Endergebnis zusammengeführt werden. Mit diesem Verfahren konnten deutliche Verbesserungen der Bildqualität und eine Stabilisierung des Verfahrens erreicht werden.

Show publication details

Henke, Birgit; Vahl, Matthias; Zhou, Zhiliang

Removing Color Cast of Underwater Images through Non-Constant Color Constancy Hypothesis

2013

ISPA 2013

International Symposium on Image and Signal Processing and Analysis (ISPA) <8, 2013, Trieste, Italy>

Color cast is a crucial problem for color image processing. White balance has been widely used to eliminate color cast to improve the image's quality. Most of white balance implementations are based on color constancy hypothesis. A wellknown color constancy hypothesis is given in [1], unifying White Patch [2], Grey World [3], Shades of Grey [4], and Grey Edge [1] assumptions in one expression. However, this general hypothesis works on underwater images not as reliable as on common images. In the color constancy hypothesis for common scenes, the ambient light source is spatial constant. But in underwater scenes, the light suffers from serious attenuation, especially in the red part of the visible spectrum. This attenuation causes spatial variance of the ambient light source, which lets classic color constancy hypothesis fail. In this paper, we propose a novel low-level image feature-based color constancy hypothesis for underwater scenes. Based on this hypothesis, we propose an algorithm, using a distance map to estimate multiple gain factors to remove the color cast.

Show publication details

Borchert, Stefan; Kröger, Willfried; Höhne, Stephan; Damaschke, Nils Andreas; Zhou, Zhiliang

On Optical Quantification of Cavitation Properties

2012

Proceedings of the 8th International Symposium on Cavitation

International Symposium on Cavitation (CAV) <8, 2012, Singapore>

This paper investigates quantitative cavitation characterization. In this context we propose shadow imaging to determine cavitation thickness and tip vortex volume. We propose a laser adjustment for absolute calibration and address cavitation extent by means of image processing. We present advantages and disadvantages of automatic processing with regard to our proposed techniques. Our main focus is on the novel cavitation thickness and tip vortex occurrence processing. Due to turbulent fluctuations all used techniques provide statistical results. The accuracy of single measurements mainly depends on camera resolution, aberrations in the optical path, illumination and optical access.

Show publication details

Gutzeit, Enrico; Vahl, Matthias; Zhou, Zhiliang; Lukas, Uwe von

Skin Cluster Tracking and Verification for Hand Gesture Recognition

2011

ISPA 2011

International Symposium on Image and Signal Processing and Analysis (ISPA) <7, 2011, Dubrovnik, Croatia>

Device free communication is getting more and more popular. A comfortable and easy way to accomplish this is interaction with hand gestures. A very good gesture recognition system is required in order to avoid frustration among users. The major challenge in hand gesture recognition lies in an accurate segmentation of the hand. The accurate segmentation of skin colored body parts is an active field of research, and much research has been done to solve this segmentation problem with skin models or skin classification rules. In this paper we address hand segmentation and introduce a novel approach based on the extraction, tracking and verification of skin clusters in color space during runtime. An evaluation shows, that our approach performs significantly better than other approaches used for comparison.