Abstarct: Most of the works about removing document nonlinear distortion is in Latin document images. The purpose of removing nonlinear distortion is compensating page curl and flattening scanned or camera captured document, so that this documents are readable by OCR. In this thesis a new algorithm is proposed for removing 3D page curl in Persian document images. In the proposed algorithm noise reduction, binarization, smearing, baxseline detection is done and then words of dewarped lines are located in the right place. In the proposed algorithm, DLCM approach is used for line detection. The output of dewarped image is presented in both binary and grayscale layout. To evaluate the operation of the proposed algorithm ,Persia negar which is a Persian OCR software is used. Experimental results show good operation of the proposed algorithm.