Images are often received in defective conditions due to poor scanning and transmitting devices. Consequently, it creates problems for the subsequent process to read and understand such images. This paper presents a novel recursive intelligent approach based on cellular neural network (CNN) to denoise an image even in the presence of very high ratio of noise. Image denoising is devised as a regression problem between the noise and signals; finally, it is solved using CNN. Accordingly, noises are detected with surrounding information and are removed. Initial experiments show that proposed approach could achieve a higher peak signal-to-noise ratio (PSNR) on images. The proposed algorithm exhibits promising results from quantitatively and qualitatively points of view.