Interpolation of Low-Resolution Images for Improved Accuracy Using an ANN Quadratic Interpolator

Abstract

The era of digital imaging has transitioned into a new one. Conversion to real-time, high-resolution images is considered vital. Interpolation is employed in order to increase the number of pixels per image, thereby enhancing spatial resolution. Interpolation's real advantage is that it can be deployed on user end devices. Despite raising the number of pixels per inch to enhances the spatial resolution, it may not improve the image's clarity, hence diminishing its quality. This strategy is designed to increase image quality by enhancing image sharpness and spatial resolution simultaneously. Proposed is an Artificial Neural Network (ANN) Quadratic Interpolator for interpolating 3-D images. This method applies Lagrange interpolating polynomial and Lagrange interpolating basis function to the parameter space using a deep neural network. The degree of the polynomial is determined by the frequency of gradient orientation events within the region of interest. By manipulating interpolation coefficients, images can be upscaled and enhanced. By mapping between low- and high-resolution images, the ANN quadratic interpolator optimizes the loss function. ANN Quadratic interpolator does a good work of reducing the amount of image artefacts that occur during the process of interpolation. The weights of the proposed ANN Quadratic interpolator are seeded by transfer learning, and the layers are trained, validated, and evaluated using a standard dataset. The proposed method outperforms a variety of cutting-edge picture interpolation algorithms..