Image Super-resolution Using Super-resolution Convolutional Neural Network


  • Nachiket Jadhav
  • Aniket Matodkar
  • Ganesh Borana
  • Sujata Bhairnallykar


Several studies have been conducted to improve low-resolution photos to a higher resolution version utilizing classic super-resolution methods such as pixel-wise interpolation. While the resolution of images has been rapidly increasing in recent years with the development of high-performance cameras, advanced image compression, and display panels, the demand to generate a high-resolution image from pre-existing low-resolution images is also increasing for rendering on high-resolution displays. Some of these methods involve cropping a large image into small parts and applying superresolution to each one individually. The downside, however, is that in many cases, the cropping lines are distinctly visible, which in turn reduces the quality of the output image. Studies concentrating on integrating CNN-based neural networks into super-resolution are evolving as deep learning technologies advance, to improve speed and minimize memory load on numerous devices. Neural networks can learn an end-to-end mapping between low- and high-resolution images by themselves. The mapping is represented by a deep convolutional neural network (CNN), which takes a lowresolution image as input and produces a high-resolution image as output.