||Study on Optical Image Quality Improvement for Underwater Environment
Since the 1960s, autonomous underwater vehicles (AUVs) and unmanned underwater vehicles (UUVs) have been used for deep-sea exploration. Sonar sensors also have been extensively used to detect and recognize objects in oceans. Although sonar sensors are suitable for long-range distance imaging, due to the principles of acoustic imaging, sonar images are low signal to noise ratio, low resolution and no colors. In order to acquire more detail information of underwater object, a short-range imaging system is required. In this situation, a photo vision sensor is used reasonably.However, the low contrast and color distortion of underwater images are still the major issues for practical applications. Therefore, this thesis will concentrate on the underwater optical images quality improvement.Although the underwater optical imaging technology has made a great progress, the recognition of underwater objects is still a challenging subject nowadays. Different from the normal images, underwater images suffer from poor visibility due to the medium scattering and light distortion. First of all, capturing good quality images in underwater circumstance is difficult, mostly due to attenuation caused by light that is reflected from a surface and is deflected and scattered by particles. Secondly, absorption substantially reduces the light energy. The random attenuation of the light mainly causes the haze appearance along with the part of the light scattered back from the water. In particular, an underwater object which 10 meters away from camera lens is almost indistinguishable because of light absorption. Furthermore, when the artificial light is employed, it can cause a distinctive footprint on the seafloor.In order to obtain high quality underwater images that can be adapted to the traditional image identification algorithms, this work aimed to construct an underwater image processing framework. Due to the special characteristic of underwater images,segment the image to several parts before directly perform a subject identification is thought an efficient way. And for obtaining a good underwater image segment result, the work to improve the quality of the image is necessary. Such work contains image enhancement, color correction and noise reduction, etc. The experiments demonstrate that the proposed methods produced visually pleasing results, and the numerical image quality assessment also proved the effectiveness of this proposal. The organization of this thesis is as follows.Chapter 1 briefly reviews the characteristics and types of acoustic imaging and optical imaging technologies in ocean. The traditional underwater imaging models and the issues of recent underwater imaging systems are also introduced.Chapter 2 describes a novel underwater image enhancement method. The transmission is estimated by the proposed dual-channel prior. Then a robust locally adaptive filter algorithm for enhancing underwater images is used. In addition, theartificial light removal method is also proposed. Compared with the traditional methods, the proposed method obtains better images.Chapter 3 presents a color correction method to recover the distorted image colors. In the experiments, the proposed method recovers the distorted colors in real-time. The color corrected images have a reasonable noise level in their dark regions, and the global contrast is also well improved.Chapter 4 describes two methods for image segmentation. The first one is the automatic clustering Weighted Fuzzy C Means (WFCM) based segmentation method. It automatically obtains a reasonable clustering result for the underwater images with simple texture. The second method is fast Active Contour Model (ACM) based image segmentation method, which dramatically improves the calculation speed. Compare with the traditional methods, the processing speed is improved by over 10 times.Chapter 5 presents the conclusions of this work, and points out some future researchdirections.
九州工業大学博士学位論文 学位記番号：工博甲第398号 学位授与年月日：平成27年9月25日
1 INTRODUCTION|2 IMAGE ENHANCEMENT|3 COLOR CORRECTION|4 IMAGE SEGMENTATION|5 CONCLUSIONS