Natural Image Interpolation Using Extreme Learning Machine

Standard image interpolation methods use a uniform interpolation filter on the entire image. To achieve improved results on specific structures, content adaptive interpolation methods have been introduced. However, these are typically limited to fit image data into a linear model in each class. In this paper, we investigate replacing the linear model by a flexible non-linear model, resulting in a novel interpolation algorithm based on extreme learning machines. Extreme learning machines (ELMs) is a relatively recent learning algorithm for single hidden layer feed-forward neural networks, which compared with conventional neural network learning algorithms, overcomes slow training speed and over-fitting problems. Based on an extensive set of experiments, we show that our proposed approach yields improved image quality, as confirmed by both objective and subjective results.