Abstract: striping in the images which acquired by the linear array ccd blocks is mainly caused by disaccordance of each ccds . This problem can be solved by relative detector calibration. How to set up a fitted model of calibration depends on the satellite itself. Although physics model is reliable, it requires delicate equipments and some useful parameters are too complicated to find out. In practice, we always use signal or statistical models based on the image itself to deal with such problems. This article developed a new statistical model :the relative calibration model based on LMS linear regression. This model has been proved to be adaptable in dealing with lots of normal remote sensing images and get good results.

Key words: LMS, relative detector calibration, striping

摘要： 卫星影像条纹噪声主要来源于ccd相机各个探元响应度的不一致，通过相对辐射校正可以解决这个问题。如何建立合适的校正模型，与卫星本身条件有很大关系。物理模型可信度强，但是要达到物理模型精度的要求，对光学和辐射的仪器要求严格，所需的有些参数过于复杂，难以得到。在条件简陋的情况下，我们也可以采用基于图象本身的统计模型。本文提出一种新的统计模型：基于最小二乘回归的相对辐射校正模型。通过实例，发现它具有适用性强的特点，可用于一般的大幅遥感图象，不仅能得到有较好视觉效果的图像，而且生成比较准确反映ccd性质的增益偏置值，适用于相同ccd在邻近时段的所成的其它图象。

关键词: 最小二乘, 相对辐射校正, 条纹噪声