Af Algorithm Of The Digital Imaging System To Study

Auto-focus is one of the key problems in computer vision,imaging systems and various precision instruments. In the last ten years,research on auto-focusing techniques has been focused on methods related to digital image processing on account of the maturing of digital imaging devices,and a lot of achievements have been accomplished. However,at present there are still many shortcomings in auto-focusing techniques used in digital imaging systems because of the complexity of imaging objects and conditions. Many problems,such as how to select focus measures and focus windows,how to perform feedback control,and how to improve the auto-focusing velocity etc.,are far from completely solved. This paper summarizes the author's research accomplishments on the project of "Auto-focusing Algorithm in Digital Imaging Systems" in the last two years.First,the paper briefs the principles and research status quo of auto-focus in digital imaging systems. Then,the author's original work is discussed in detail:(1) Focus measure based on RGB analysis is proposed to process color images where gray levels are almost uniform and the traditional approach fails. (2) Focus measure based on wavelet analysis is introduced to differentiate those images taken slightly defocused;(3) A simple and functional optical imaging model is presented. On the basis of the model the necessity to select focus windows is expatiated,and the limits of the traditional methods are shown. Then the general strategies to choose focus windows are brought forward. (4) Pattern recognition methods are ushered into auto-focus. Selecting skin areas in images as focus windows is put forward for images where imaging objects are men. To realize this novel method,a simple color space converting is performed to distinct skin pixels from non-skin ones,which utilizes the characteristics of human skin color distribution. After that,morphological operations are used. Similar approach is possible for images where imaging objects are easy to be extracted. (5) Tracking pupils of photographers to select focus windows is proposed for images where it's difficult to extract .imaging objects. The following problems are preliminarily solved:determining the relations between the positions of one's pupils and his interested areas;modifying the architecture of current imaging systems to take images of photographers' eyes;deciding the positions of pupils by processing the eyes' images taken. (6) The feedback control hi auto-focus and the hardware and software architecture are also discussed.