The invention discloses a method for removing cloud noise effects in a normalized difference vegetation index (NDVI) time sequence image. The method comprises the following steps of: S1, generating an annual NDVI time sequence image according to NDVI data of every ten days; S2, carrying out least square fitting on all time sequence pixels in the NDVI time sequence image, and assigning same values to weights of all points relative to a curve; S3, comparing an observed value with a fitted value, and eliminating points under negative cloud action; S4, repeating the steps S2 and S3, eliminating all the points under the negative cloud action, and generating a new curve; and S5, post-processing the curve obtained in the S4 to obtain a finally fitted curve. In the method disclosed by the invention, by utilizing an NDVI time sequence file and a harmonic function analysis method based on the least square fitting, cloud shielding and atmospheric effects on a sensor in a data acquisition process are effectively removed, the time sequence image with cloud contamination removed can be generated, the obtained curve has the advantages of obvious trend, strong relative property among years and high precision, and the method has a wide applicable range.