<p>Today’s digital world relies heavily on images from drones, satellites, smartphones, and medical scanners, which are often corrupted by noise due to environmental conditions, hardware limitations, and data compression. Noise degrades image quality and interferes with tasks such as object detection, facial recognition, and medical diagnosis. However, current denoising algorithms often assume that the noise type and level are known in advance, which leads to poor performance when noise characteristics are unknown. To overcome this limitation, this paper proposed a new hybrid method that automatically determines the type of noise and measures its level before applying the appropriate denoising technique. While many deep learning models can classify noise types or perform denoising. This paper proposes noise type identification and estimates the noise level, enabling us to choose the appropriate method for denoising a particular noise type. We develop a two-stage model, comprising an enhanced CNN for identifying nine types of noise and a Hybrid deep learning model to quantify noise levels from 0 to 90 percent. Proposed models were trained and evaluated on CIFAR-10, MNIST, and BSD500, and the Landscape grayscale image and the BSD500 datasets were used for noise level estimation. Experimental results illustrated the effectiveness of the Enhanced CNN model and a hybrid deep learning model in precise noise identification and estimation.</p>