High dynamic range (HDR) imaging technology has been widely implemented in digital microscopes for taking still images of high-contrast specimens. However, capturing HDR microscopic video is much more challenging. In this dissertation, an HDR microscopic video system based on GPU accelerated computing is presented. By combining CPU and GPU computing, it is possible to build a stable HDR video system using a single off-the-shelf camera. The computing efficiency analysis shows that capturing multiple frames of different exposure intervals, aligning consecutive neighbouring frames, constructing HDR radiance map and tone mapping the radiance map for display, can all be realised by using GPU computing to accelerate the processing speed. The experimental results were presented to show the effectiveness of the system and how HDR video can reveal much more detail than conventional videos.
The idea of employing HDR imaging technology in 3D surface construction has been proposed as a solution to the Shape From Focus limitation. Shape From Focus (SFF) is the most effective technique for recovering 3D object shape in optical microscopic scenes. Although numerous methods have recently been proposed, less attention has been paid to the quality of source images, which directly affects the accuracy of 3D shape recovery. One of the critical factors impacting source image quality is the high dynamic range issue, which is caused by the gap between the high dynamic ranges of the real world scenes and the low dynamic range images that the cameras capture. To overcome this issue, a novel microscopic 3D shape recovery system based on high dynamic range (HDR) imaging technique is proposed. By combining SFF and HDR, it is possible to build a robust 3D system using a single off-the-shelf camera and a traditional optical microscope. Experiments on constructing 3D shapes of difficult-to-image materials have been conducted, in terms of metal and shining plastic surfaces where conventional imaging techniques will have difficulty capturing detail, and will thus result in poor 3D reconstruction. The experimental results show the proposed HDR-based SFF 3D method yields more accurate and robust results than traditional non-HDR techniques for a variety materials.
After the analysis of HDR and Shape From Focus techniques, another project about microscopy was presented, which is tuberculosis bacteria detection. Tuberculosis (TB) is an infectious disease in low- and middle-income countries. There are many tools behind physical examinations for TB detection, but the most effective method is visual examination using microscopes, in terms of fluorescent microscopy and bright field microscopy. However, the former method is on average 10% more sensitive than the latter. This project not only aims to detect tuberculosis automatically to help technicians, but also aims at the construction of a subsequent autofocus system based on the detection of tuberculosis. The focus analysis, which is the initial step of shape from the focus technique, acted on the region of tuberculosis exists, regardless of the other areas. In this case, a new TB detection method based on Random Forest using fluorescent microscopic images was presented. Experiments on three types of classifiers, in terms of Random Forest (RF), linear SVM (LinSVM), Cross-Validation SVM (CVSVM), were conducted. The experimental results indicate that the RF-based learning method for TB bacteria classification using fluorescent images achieved higher performance than the other two machine learning methods.
|Date of Award||10 May 2017|
- Univerisity of Nottingham
|Supervisor||Guoping Qiu (Supervisor) & Rong Qu (Supervisor)|
- 3D construction
- Tuberculosis detection