漸進背景影像的建構

Progressive Background Image Generation

在對靜態攝影機取得連續影像處理時，為了要能快速地偵測出影像中的前景物體，較簡單的方法，就是事先建構好背景影像，之後將連續輸入的影像和背景影像相減，即可快速地取得前景物。而建立背景影像時，若能事先淨空環境，則背景影像的建立並不困難；但是在某些運用，如交通監控，事先環境淨空並不可行。本文提供一種演算法，可以從非淨空的環境視訊中，漸進式地建立起環境的背景影像，又由於這種漸進式的機制，背景影像會隨著環境的變化，隨時自動地更新。過去的方法常將建構與更新的步驟分開，本文所提的方法則將兩者結合成一體，且可同時應用在單色及彩色影像上，目前實驗的結果相當符合人類的認知。

Background images play an important role in the rapid extraction of foreground objects from their background scene. In this paper, a new technique of background image generation via the use of partial background images is presented. Partial background images tare helpful in facilitating the progressive process of image extraction and may also eliminate the unstable effects of the camera&apos；s physical vibrations. In this method, histograms are utilized to record and trace the changes in intensity in order to determine the best background values. This procedure does not require a large space in which to store the image sequences, and it is fast since the constructing and updating processes can be performed simultaneously. Moreover, variations in illumination or in the weather may be tolerated during background image generation, and the method is applicable to both gray and color images. The feasibility of the proposed technique is strongly suggested by our experimental results, which are reported at the end of this paper.