適用於低色彩對比前景抽取之CIELAB色彩空間背景模型

Abstract

利用固定攝影機拍攝的串流影像資訊於前景物體抽取是一個很典型的方法。在一般前、後景色彩深淺差別大時，可以簡單的使用亮度的資訊將前後景分離，但當前後景色彩接近時，例如; 當辨識的目標穿著和背景相似的衣服時，若只使用灰階影像並無法將完整的前景資訊分離，我們曾使用HSV色彩空間加入像素點色彩成分的考慮建立背景模型做顏色的補償，達到前、後景的分離，且能對陰影的問題加以消除改進。然而使用HSV色彩空間會遇到色調一些不穩定的問題，所以我們在色調不穩定的區域加以限制，以增加抽取前景影像的準確性，但對於某些情況，例如;背景為米白色而前景目標穿著粉紅色衣服時，在HSI系統對前景物體抽取的準確性提升效果有限。 本論文，我們建立一個內嵌在CIELAB色彩空間的統計性背景模型來做前景物體抽取，這個模型大幅的提高前景物體抽取的靈敏度。在HSV的系統與我們新的前景抽取系統比較，實驗證明，CIELAB其正確率從原來的75.62%改善為87.88%。

Background subtraction is a typical method used to extract foreground object in video streams taken from a static camera. When the foreground color is different from the background color, the foreground subject can be extracted easily by the luminance component. When the foreground color is similar to the background color, we cannot extract the foreground image completely by the luminance component. To solve this, we used to utilize the HSV color space to build the background model to do color compensation, in line with similar spirit of W4 segmentation algorithm. This approach can not only extract foreground image well but also be helpful to shadow removal. However, H and S components are not consistently reliable in some situations. For example, HSI system does not detect foreground well when the object wears pink clothes when in ivory background. In this thesis, we build a statistical background modeling embedded in CIELAB color space for foreground object extraction. By the use of color difference formula in CIELAB space, so that the sensitivity of foreground object extraction can be raised evidently. In comparison with HSV based scheme and our new foreground extraction scheme, the CIELAB improves the segmentation accuracy from 75.62% to 87.88%.