Single Color Image Super-resolution through Neighbor Embedding
Mei Gong, Kun He, Jiliu Zhou, Jian Zhang · 2011
Quaternion is firstly adopted to model a color image directly, and based on the quaternion model, a single color image super-resolution algorithm through neighbor embedding is proposed. Given a low-resolution color image, its high-resolution color image is reconstructed from a set of training images, which are composed of one or more low-resolution and corresponding high-resolution color image pairs. When doing neighbor embedding, quaternion is used to deal with the R, G, B components simultaneously. Simulation results show that the correlations of the three components are fully used in the proposed algorithm, so that the complete information of high-resolution image can be effectively reconstructed.