Image Edge Detection Based on a Spatial General Autoregressive Model

A new scheme for image edge detection based spatial general autoregressive model (SGAR) is proposed in this work, which takes into consideration the nonlinearity at both edges and non-edges. First the SGAR model is derived which fuses both linear and non-linear spatial autoregressive model in one expression and whereafter the spatial relation of a lattice site with its neighbors is adaptively learnt by using a SGAR model with a robust parameter estimator named GM estimator. Then the edge image is produced by thresholding the residual image which is equal to the difference between the original image and the prediction image. Finally, experiments are carried out on a worldwide dataset to verify the feasibility of the proposed scheme. Experiment results also indicate that the future works need to be carried out to further improve practicability of the proposed scheme.