一種新式圖像特徵－色彩均勻區塊

Abstract

圖像特徵被廣泛的利用在影像處理及電腦視覺領域，且已有許多圖像特徵被用來解決不同的圖片相關之問題。然而現存的特徵大多為基於圖像的前景資訊。在本論文中，基於利用圖像背景相關之資訊而非前景資訊，我們提出了一種新穎的特徵-色彩均勻區塊。為了凸顯此特徵之優點並解釋此特徵之擷取、匹配過程，我們將此特徵應用於兩個領域上: 工業監控視覺化系統之主視窗辨識及台幣十元硬幣辨識。實驗結果顯示我們所提出的此兩種方法皆達到高準確率且高效率，顯現出我們所提出的特徵皆適用於此兩個領域。

Image features are widely used in image-processing and computer vision. There are several image features have been used to address various image-related problems. However, most of the existing image features are based on the foreground information. Instead using the foreground information, we introduce a novel feature called uniform block, which utilizes the background information. To outline the advantages and illustrate the feature extraction and feature matching of uniform blocks, we apply the uniform block feature on two subjects: top window recognition of industrial visualization systems and NT$10 coin recognition. Both of the experimental results have reached high accuracy and perform efficiently, indicates that the proposed feature is suitable for these subjects.