一种基于LASSO回归模型的彩色相机颜色校正方法

彩色相机的颜色校正是实现成像色彩一致性的必要保障手段。传统的相机颜色校正中,对测量数据多采用多项式回归分析来确定颜色定标系数,存在着精度不高的缺点,因此,本文对测量数据提出了基于LASSO的高阶多项式回归拟合方法,利用LASSO压缩系数的特点,在保证计算复杂度的前提下,有效提高了回归模型的校正精度。在D65标准光源下对ColorChecker 24色卡进行了实际成像实验,并用CIELAB色差公式表征了校正效果,实验结果表明,新方法的校正效果明显优于传统的线性回归、二次多项式回归方法,平均色差指标可以达到5个CIELAB色差单位。

A Color Correction Method of Color Camera Based on LASSO Regression Model

Color correction is the essential safeguard method to keep the image color constancy for a color camera.In traditional color correction methods,the color calibration coefficients are mostly obtained by the polynomial regression model,whose correction accuracy is often insufficient.So a high order polynomial fitting method based on LASSO (Least Absolute Shrinkage and Selection Operator) regression model is proposed,considering that LASSO can compress the polynomial coefficients efficiently to guarantee the model complexity and improve the correction precision as well.The experiments are conducted using D65 standard light source and ColorChecker 24 as the imaging object.The results characterized by CIELAB color difference function show that compared with traditional linear regression and quadratic polynomial regression methods,the images corrected by LASSO method has a better effect,with the mean color difference value about 5.