Adaptive ANN-based method of constructing an interpolation formula for doubling the image size

Abstract

The architecture of an artificial neural network that solves the problem of constructing interpolation formulas for doubling the size of images is proposed. The trained model receives a 4×4 matrix as an argument. The result is an interpolation formula represented as a weight vector for 4 points. A comparison of the main quality assessments of the proposed method with some well-known adaptive approaches is made. The results of the comparative analysis show that the proposed approach has a better interpolation quality than NEDI and DCCI methods.

Keywords

• interpolation
• machine learning
• artificial neural network
• gradient descent
• image quality