同步式多重亮度紅外線打光器之校正方法

Abstract

夜間監控攝影因為無足夠的環境自然光源，所以需使用打光器來增加環境光源，但由於打光器的光源強度及放置位置的不同，常會造成影像的曝光結果有過暗或過亮的問題，使得影像品質無法達到人類感官滿意的範圍。使用多重亮度紅外線打光器能產生多個不同光源強度來得到多種不同曝光值的影像，而後可以再從這些影像挑選出曝光較佳的影像，使影像品質獲得改善。本論文提出自動校正的方法，其主要目的在於校正同步改變亮度之紅外線打光器，使校正過後的打光器所產生之亮度差異，對於在不同的環境下所拍攝到的不同景物影像，均能有效涵蓋具有最佳的辨識度之影像。如此，可以減少因設定不當的打光強度所產生過暗或過亮的影像數量，進而增加各亮度所提供的有效資訊，而更有機會對於所挑選各亮度影像能獲得有用的影像分析結果。實驗結果顯示，本論文所提出之自動校正的方法，確實在人臉偵測方面能達到上述之效能提升效果，而其可能應用還包括車牌偵測、高動態範圍影像合成…等。

In nighttime surveillance, there is usually not enough natural ambient light for image capturing, and an IR illuminator is often used to partially resolve the problem. However, improper intensity/location of an illuminator may result in object images which are too dark or bright so that image quality is not satisfactory even for human perception. Using multiple-intensity IR-illuminator can produce a number of different light intensities to help the generation of images with different degrees of exposure, from which well-exposed images can be chosen, and increase the number of good quality images. The goal of this thesis is to develop an automatic calibration method for a synchronous multiple-intensity IR-illuminator, which may be used in different environments for capturing images of objects at various locations, so that quality images, e.g., those with good contrast, will be included in the images captured with selected illumination intensities. In other words, by reducing the number of improperly illuminated images, more informative images (and more useful image analysis results) can be obtained. Experimental results shows that the proposed automatic calibration method can indeed achieve the above improvement for the performance of automatic face detection. Potential applications which may also benefit from such calibration include license plate identification, image integration for a high dynamic range (HDR) representation …etc.