Striping noise is the common system noise during formation of image using linear array CCD and has the character of periodicity, directivity and banding distributing. It can be caused by errors in internal calibration devices, or by slight gain/offset differences among the elements that conform the array of detectors. Striping noise covers up useful information in CCD image and brings adverse effect to image interpretation. On the basis of analyzing wavelet decomposed coefficient, the regularities of distribution about striping noise in wavelet coefficient is found, thereby the method of wavelet threshold selection which is suitable to striping noise distribution is put forward. According to Donoho's method about denoising using wavelet, the image including striping noise is processed. Comparing the power spectrum of processed image with the one of original image polluted by striping noise in frequency field, we find pulse brought by striping noise is removed and the goal which reserves image details and reduces stripes is achieved.