Image processing is a very popular and widely used technique. In this paper, we propose a framework for realizing reversible image processing, which enables that the users can return the processed image to the original copy without loss. In the proposed method, the original image is firstly processed to get the desired target image by a classic image processing method. Then the original image is reversibly processed according to the transition probability matrix, getting the processed image similar to the target image. We take histogram equalization and gamma transform as examples to show that the proposed method can realize reversible image processing and achieve nearly the same processing effect as done by irreversible image processing tools.