摘要
颜色直方图是图像检索技术中一种有效的颜色表达方式,具有位移、旋转不变的特性。然而木材表面颜色较为单一,采用通常的量化方案不能达到很好的分类识别效果。提出了采用HSV颜色空间三个独立分量的直方图统计特征表达木材表面颜色信息的方法,利用色调、饱和度和亮度三分量之间的相互独立性,提取了各分量的直方图特征。最后利用BP神经网络对木材样本库的图像进行了分类仿真,其结果验证了特征的有效性。
Color histogram is an effective color expression in Content-based Image Retrieval (CBIR), which possesses the invariability in displacement and circumrotation. As color of wood surface is rather singleness, the usual quantification can not get a satisfied classification result. The statistical features of color histogram of three independence variables in HSV color space, and independent characteristics of hue, saturation, and brightness are adopted to express the information of color of wood surface in this paper. At last classification simulation is done by using BP neural network, and the results of which testify the validity of features.
出处
《森林工程》
2008年第1期34-36,共3页
Forest Engineering
基金
哈尔滨市自然科学基金项目(2004AFXXJ020)
黑龙江省自然科学基金项目(C2004-03)资助