Abstract: Image denoising is a challenging task and defined as an inverse problem of estimating a signal from the noisy measured version. (Vo and Le Thi, 2015) has presented a DC algorithm for the image denoising problem using sparse representation and dictionary learning when approximated l0-norm by capped-l1function (Peleg and Meir, 2008). In this report, we will study the Exponential function (Bradley, 1998) to approximate l0- norm and propose a DC algorithm for this new approximation problem. The algorithm will be applied in denoising problem. Experimental results on the real data sets are promising and demonstrate the effectiveness of our approach.