A new image color feature extraction method
Qiu Zhao-wen, Tian Zhang · Ha'erbin gongye daxue xuebao · 2004
A new 20 color none uniform quantization algorithm based on HSV space was proposed with the purpose of reducing the dimension of feature space and syncretizing the transcendental knowledge into the classification machine constructed by SVM. Comparing with the traditional color quantization algorithms, the complexity of time and space decrease and the retrieval accuracy is improved. The transcendental information can be syncretized into the nucleus function of SVM to improve the effect of classification. Experiments show that the new algorithm proposed can be successfully used in retrieving the image from good-sized image database and extracting semantic information from image automatically.