Windowed fourier transform based fringe pattern analysis techniques

• Main Author: Zhao, Ming
• Other Authors: Qian Kemao
• Format: Theses and Dissertations
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/68569
• Institution: Nanyang Technological University

Measurement techniques are powerful tools for obtaining properties of objects and recording their changes under different conditions. The demand of those valuable information promotes the development of measurement techniques. Fringe pattern based techniques are an important family of measurement techniques. They cover a measurement range from nanometers to kilometers. Due to the importance of fringe patterns, fringe pattern analysis techniques are widely used and studied in academia and industry. However, as the requirements of measurement techniques are increasing, people still need to seek more reliable, more accurate and faster analysis techniques. Windowed Fourier transform (WFT) algorithms, including windowed Fourier ridges (WFR2) and windowed Fourier filtering (WFF2) methods, are effective and automatic fringe pattern analysis techniques. They can be applied for fringe pattern denoising, demodulation and providing local fringe properties for other analyses. One drawback of the WFR2/WFF2 algorithms is that they are very time-consuming. In this dissertation, factors for an efficient implementation, such as the selections of Fourier transform libraries and parallel techniques, are explored in detail, and a multi-threaded implementation is presented. Fringe patterns are special images, so they can be denoised by general image denosing methods. To understand the performance of general image denoising methods on fringe patterns, the block-matching and 3D filtering (BM3D) method is selected and compared with the WFF2 method. The BM3D method does not outperform the WFF2 method in general, but its performance in discontinuous areas is better than the WFF2 method. This finding inspires us to combine them together. The hybrid method is proposed and it achieves better denoising results than using those two algorithms individually. Besides fringe pattern denoising, phase unwrapping is another important task of fringe pattern analysis. In this dissertation, quality guided phase unwrapping (QGPU) is reviewed by comparisons of quality maps and guiding strategies. New data structures are also proposed for accelerating the QGPU process. The QGPU process is vulnerable to discontinuity. A snake assisted QGPU (sQGPU) method is proposed to solve the discontinuity problem. With the aid of the snake model, piece-wise phase unwrapping, which is immune to the discontinuity problem, is finally achieved.