论文部分内容阅读
由于数码相机的颜色空间是依赖于设备的,对于一个具体的数码相机,其光谱响应与设备独立的CIE标准观察者颜色匹配函数是一个非线性关系,因此不能真实复制场景的颜色。特性化彩色图像设备是提高图像的颜色复制质量的一个重要方法。介绍一种基于BP神经网络数码相机特性化方法。采用Munsell颜色系统作为目标色,大样本训练空间。测试了不同的网络结构和样本空间分布。训练样本平均色差为1.75CMC(1∶1)色差单位,测试样本为2.16。该方法在数码相机颜色测量、光谱重建等领域有广泛的应用前景。
Because the digital camera’s color space is device dependent, the spectral response of a particular digital camera is non-linear with the device-independent CIE standard observer color matching function and therefore can not truly duplicate the scene’s color. Characterizing color image equipment is an important way to improve the color reproduction quality of an image. This paper introduces a digital camera characterization method based on BP neural network. Munsell color system as the target color, large sample training space. Different network structures and sample space distributions were tested. The average color difference of training samples is 1.75CMC (1: 1) color difference unit, the test sample is 2.16. The method has wide application prospects in the field of digital camera color measurement and spectral reconstruction.