A Study of Image Processing Techniques for Image Deblurring and Automated Optical Inspection

• Main Authors: Chia-Feng Chang, 張嘉峰
• Other Authors: Jiunn-Lin Wu
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/73804006758798998344

博士 === 國立中興大學 === 資訊科學與工程學系 === 105 === The goal of image processing technique is to process the image to be readable by human being or computer. Image deblurring is a fundamental and challenging problem in both image processing and photography with broad applications. Under low light conditions, the use of a longer exposure time can provide a brighter image but may result in motion blur due to the camera shaking. A blurred image can be modeled as an unblurred image convolutes the movement of camera (or called the point spread function, PSF) if the point spread function of motion blur is assumed shift-invariant. If we apply an inaccurate PSF to restore the blurred image, the ringing artifacts may appear in the deblurred image. The ambiguity problem and enhanced noise are the main causes of an inaccurate PSF estimation. To address these problems, we proposed several image deblurring algorithms. To reduce the ringing artifacts, which were caused by an inaccurate PSF, we propose applying an adaptive edge map to image deblurring. For the ambiguity problem, we propose an improved patch based sharpening method to avoid this problem. However, the shift-invariant assumption does not always hold, we propose uniform defocus map to reduce the shift-variant problem into shift-invariant problems. Image deblurring is an image restoration technique, which recover an unblurred image from blurred image. It can apply to many applications to help to obtain a better performance, for example the automated optical inspection. Automated optical inspection (AOI) is an automated visual inspection of industrial product where a camera autonomously scans the device under test for both catastrophic failure and quality defects. The procedure is the camera captured the digital image of production on the production line, and analyzed the captured digital image by defect detection algorithm for quality inspection. The difficulty of the defect detection is the position and shape of defect is variety on different production, therefore there is no universal defect detection algorithm for detecting all kinds of defects on various productions. We have to design the specific defect detection algorithm to different production. In this study, we design hybrid based inspection methods for both wafer chip and micro compact camera lens. The captured image of wafer chip has complicated textures include defect, particle, reflection and dust. We detect each type of target separately; as a result, we can get a high accuracy detection result. To detect the defect on the camera lens is a hard work because of the circular texture. We apply polar coordinator transform to transform the circular region into rectangular region, therefore we can avoid the influence of circular texture and defect the real defects. However, during capturing image, a blur artifact was generated in the captured image because the ambient light is not fine-tuning or the focal length of camera is not focus on production. The blur problem reduces the contrast of defect so that we cannot detect the defects in the image. We apply the image deblurring solve the blurred image problem. Experimental results present that proposed method can achieve satisfactory results.