A preliminary study of ANN implementing image filters
A. Jhih-Hao Chen, B. Chih-Wen Wang, C. Jyh-Horng Jeng · 2013
Artificial neural network (ANN) is a learning machine possessing the universal approximation property which can approximate mathematical models to a pre-specified degree. In this paper, we use ANN to implement average, Sobel, and Laplacian filters for image processing. This paper is a preliminary study of applying machine learning to image processing. For future study, it can be extended to more complicated image processing such as super-resolution and compression by using sophisticated machine learning techniques. Experiments are conducted to test the approximation ability of ANN with global and local training sets.