| dc.description.abstract | 在本論文中，我們發展出顯微照像的快速自動對焦技術，主要以解決目前在TFT-LCD及彩色濾光片 (color filter, CF) 面板產業所使用的自動光學檢測機 (automatic optical inspection, AOI) 及雷射修補機 (laser repair, LR) 所需的自動對焦問題為基礎，提供ㄧ個快速、穩定、且可靠的顯微照像快速自動對焦技術。
我們依據光學成像原理以及影像處理的概念，使用快速傅立葉轉換 (Fast Fourier Transform, FFT)、離散餘弦轉換 (Discrete Cosine Transform, DCT)、小波轉換 (Discrete Wavelet Transform, DWT)、及理想高通濾波器 (Ideal high pass filter, IHPF)、巴特沃斯高通濾波器 (Butterworth high pass filter, BHPF)、高斯高通濾波器 (Gaussian high pass filter, GHPF)、理想低通濾波器 (Ideal low pass filter, ILPF)、巴特沃斯低通濾波器 (Butterworth low pass filter, BLPF)、高斯低通濾波器 (Gaussian low pass filter, GLPF)等頻率域的濾波器、及空間域的濾波器ㄧ次差分平方和 (energy of gradient magnitude, EOGM) 與二次差分平方和 (energy of image Laplacian, EOIL) 等技術來分析顯微影像的清晰程度。並將上述技術結合二次曲線近似、三次曲線近似、與高斯曲線近似等對焦搜尋演算法 (autofocus searching algorithm)，探討其移動間距、取樣點數、對焦速度、及對焦穩定性與可靠度。為了達到快速、穩定與可靠等特性，我們提出二種判定對焦量測準則優劣的準則：(i) 雜訊的抵抗能力；(ii) 計算複雜度；並提出二種判定對焦搜尋演算法優劣的準則：(i) 移動取像次數；(ii) 正確對焦的穩定性與可靠度。我們認為對焦量測準則與對焦搜尋演算法有顯著的相關性，因此我們依據對焦量測準則的雜訊抵抗能力、計算複雜度、及對焦搜尋演算法所需的移動取像次數來比較對焦量測準則；最後以大量的自動對焦實驗來證實，我們發展出的顯微照像快速自動對焦技術可以達到快速、穩定與可靠的結果。 | zh_TW |
| dc.description.abstract | In this thesis, we develop a fast autofocus technique for microscope imaging, In order to solve the problem of focusing automatically in automatic optics inspection machine (AOI) and laser repair machine (LR) that used in TFT-LCD and color filter panel industry, we will offer a fast, steady, and reliable autofocus technique.
We follow the principle of the geometric optics model of image formation and the concept of the image processing, use Fast Fourier Transform (FFT), Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), frequency domain filter (Ideal high pass filter, IHPF),(Butterworth high pass filter, BHPF),(Gaussian high pass filter, GHPF),(Ideal low pass filter, ILPF),(Butterworth low pass filter, BLPF),(Gaussian low pass filter, GLPF), and spacial domain filter (energy of gradient magnitude, EOGM),(energy of image Laplacian, EOIL) technologies to analyze the focusing degree of the image of microscopes. Also, we combine above-mentioned technologies to the quadratic curve fitting approximate, the cubic curve fitting approximate and gaussian curve fitting approximate etc, autofocus searching algorithm. Probe into the moving step, sampling times, focus speed, and focus stability and reliability.
In order to reach characteristics such as fast, steady and reliable, we propose two kinds of good and bad criteria of judging focusing criterion: (i) Ability of resisting of the noise; (ii) Calculating complexity; Propose two kinds of good and bad criteria of judging autofocus searching algorithm: (i) Move and sampling times; (ii) Stability and reliability of the correct focusing. We think the focus criterion has apparent dependence to autofocus searching algorithm, therefore we compare the focus criterion base on the ability of resisting of the noise, calculating complexity, and the sampling times of autofocus searching algorithm. Finally, we use a large number of focus experiment to prove we develop a fast autofocus technique for microscope imaging, which can reach fast, steady and reliable results. | en_US |