| 摘要: | 本論文中，首先分析影像中的色彩，然後把其結果應用到色彩量化、影像搜尋及人臉偵測上。我們首先我們提出了一個色彩量化的獨創方法，此法的量化誤差非常的小，這歸功於我們結合於精減RGB色彩空間及適應性群聚方法，並且結合指定色點分類法也可加速執行速度。我們提出的方法不但有好性能且執行快速。另外，我們介紹以色彩區域為基礎的多專家影像搜尋系統，在此系統中我們採用了三個互補的色彩區域為基礎影像搜尋系統；色彩影像搜尋系統、形狀影像搜尋系統及相對關係影像搜尋系統。我們先定義虛擬機率來表示雙個影像的相似性，再利用測量值相依方法來融合各別搜尋系統使成總成相似機率，根據此總成相似機率，我們可以把相似影像從影像資料庫中依序找出來。除此之外，我們尚且利用動態選擇方法來提昇整體的性能，從實驗中可以証實本方法的可行性。 最後，我提出了一個結合色彩及形狀的方法來偵測人臉位置。利用群聚為基礎的分割方法，我們可以儘可能放寬色彩頻帶來包含所有人臉，如此便可以從各種複雜背景影像中正確偵測出人臉正確位置所在，而不必限定於單純背景或需打燈光，實驗結果也很令人滿意。 The colors embedded in an image are firstly analyzed. Then, the results are applied to color quantization, color image retrieval and face detection. In the dissertation, an adaptive clustering algorithm for color image quantization is presented first. In our approach, a superposed 3D histogram is first calculated. Then, the sorted histogram list is fed into an adaptive clustering algorithm to extract the palette colors in the image. Finally, a destined pixel mapping algorithm is applied to classify pixels into their corresponding palette colors. The quantized error of our proposed algorithm is very small due to the combination of the reduced RGB color space utilization and the adaptive clustering algorithm. Besides, the executing speed of our proposed algorithm is also quite fast due to the reduced RGB color space, sorted histogram list, suitable color design and destined pixel mapping. Experimental results reveal the feasibility and superiority of our proposed approach in solving color quantization problem. Secondly, a novel region-based multiple classifier color image retrieval system is presented. In our approach, a region-growing technique is firstly employed to cluster connected color pixels with the same color in an image to form color regions which are the primitive elements utilized in our proposed approach. Then, three complementary region-based classifiers are selected in the classifier selection stage, which include color classifier, shape classifier and relational classifier. In each classifier, a virtual probability representing the probability that an image is similar to the query image is defined. Thereafter, a set of virtual probabilities is calculated in each classifier. Next, the measurement dependent methods are applied to combine the virtual probabilities of classifiers in the decision combination stage. Besides, the dynamic selection scheme designed in the decision combination stage can further improve the system performance dramatically. Experimental results further reveal the feasibility and validity of our proposed approach in solving color image retrieval problem. Lastly, a novel face detection algorithm is presented to locate multiple faces in color scenery images. A binary skin color map is first obtained in the color analysis stage. Then, color regions corresponding to the facial and non-facial areas in the color map are separated with a clustering-based splitting algorithm. Thereafter, an elliptic face model is devised to crop the real human faces through the shape location procedure. Last, local thresholding technique and a statistic-based verification procedure are utilized to confirm the human faces. The proposed detection algorithm combines both the color and shape properties of faces. In this work, the color span of human face can be expanded as wilder as possible to cover different faces by using the clustering-based splitting algorithm. Experimental results also reveal the feasibility of our proposed approach in solving face detection problem. |
