Noise classification plays a critical role in the preprocessing phase of computer vision and image processing. It is essential for accurately distinguishing noise from meaningful image details, ensuring that subsequent enhancement and analysis tasks are effective and reliable. This step is aimed at the enhancement of image for better analysis and interpretation. Images can be noisy due to various factors like variation in brightness, sensor limitations or during transmission and acquisition of digital images. There are many conventional and deep learning approaches available to classify the noises; In this paper, we have developed a novel hybrid deep learning model by integrating existing approaches to accurately quantify impulse noise in digital images. Essentially, the hybrid model learns the statistical patterns introduced by noise and classifies the intensity of noise based on those patterns. Compared with the existing deep learning models the hybrid model performs better in classification of noise levels in an image.