In pattern recognition, we often use color histograms to detect objects, but it is not easy to recognize the same objects with different color styles, because these objects have in general different color features. Therefore it is necessary to conduct the color style classification. We raise an effective method to classify video color styles by cumulative histogram saturation (S) in HSI color space. Through training a large number of videos, the saturation (S) component of the cumulative histogram of training videos can be obtained. Two important parameters, i. e. , mean and variance, can be acquired by using matrix analysis. We try to find the different saturation range of videos of different color styles. By testing 200 video segments, for most of the videos, the result of color style classification is accurate and comprehensive, and the accuracy rate reaches 85％ or more.%在进行图像模式识别时,经常利用颜色直方图进行目标识别,由于不同色彩风格中的相同目标往往会有不同的颜色特征,有必要对其进行色彩风格分类.由此提出一种利用HSI颜色空间中饱和度(S)的累计直方图进行视频色彩风格分类的方法.通过对大量视频进行训练,得到训练视频的饱和度(S)分量的累计直方图,利用矩分析法得到直方图的均值和方差,统计得到不同视频色彩风格所在区间范围.通过对200段视频进行测试,大多数视频色彩风格分类的结果是准确、全面的,准确率达到了85%以上.
展开▼
