An Approach for Determining Angle of Rotation of a Gray Image Using Weighted Statistical Regression

Abstract— This article presents a computationally efficient method to estimate two dimensional in place rotation. This method is based on the estimation of weighted linear regression lines of the reference and the sense image pair. Angle between these two regression lines can identify the rotational angle between the reference and sense images. This fully unsupervised technique offers convincing result for wide range of images, without any need for setting control points. Another important contribution is that in composite rotation this method can effectively identify cumulative angle of rotation. The experiments show that the proposed method is robust and can be applied on various kind of image applications. Comparison with existing methods suggests the supremacy of the present one over its existing competitors.