The purpose of this research work is to develop a system to correct the noise present in images, using computer vision and automatic learning tools, specifically, focused on processing images used by assisted driving systems, in which the presence of noise represents a loss of information, directly affecting the efficiency of such systems.To achieve the above, a methodological process of analysis, design, and evaluation was followed, thus obtaining a neural network model capable of responding to the proposed task, which was verified by performing tests and applying metrics such as PSNR and SSIM, observing that the system can increase up to 14 dB the PSNR value and approximately 0.76 the SSIM value concerning the initial one, this specifically for the case of the tests with Gaussian noise, while for the Salt & Pepper noise the PSNR value increased by 8.53 dB and approximately 0.73 the SSIM value concerning the initial one.