Experimental Proof Manifest Bilateral Blur as Outstanding Blurring Technique for CNN Based SR Models to Converge Quickly

Image super-resolution is utilized in a variety of applications, including medical imaging, video surveillance, satelite photography and astronomy, fraud detection, real-time scene monitoring for security and others. The most recent literature encompasses a number of image low- resolution methodologies. Image processing based on CNN deep learning’s super-resolution techniques are the simplest and most resource-efficient. In contrast, the original SRCNN architecture incorporated Gaussian blur, which takes a long time to converge. This not only increases the computational cost, but it also extends the training time. As a result, we retrained SRCNN using widely available blurring and discovered that the bilateral filter outperforms the Gaussian filter. Our objective is to keep the original SRCNN functionalities while shortening training time and computational requirements. To the best of our knowledge, this is the first research of its sort, and no other study has applied this alternative blurring training on SRCNN and ranked it as the best for image super-resolution.