Image invariant via the Radon Transform

Proposes an alternative approach based on a new image invariant computed from the Radon Transform. The authors recall the definition of the Radon Transform which is a basic operator in the field of image reconstruction from projections. The Radon Transform associates to a 2D image the set of its integral on straight lines. It can be parametered by two variables (space and angle), and can be represented as a new 2D image. They then derive its properties relatively to translation and rotation. More precisely both operations are traduced by a translation on the Radon Transform image. These properties are illustrated by numerical examples. They propose a spectral image invariant computed from the Radon Transform. The relation between this quantity and the 2D Fourier Transform of the image is given. They then discuss the advantages and drawbacks of the computation of the image invariant by each of the two methods. At last some results computed on simulated digital images are presented.<
>