Information-based shape description with scale, translation and rotation invariance

This paper proposes a technique based on information theory principles for shape description of 2-D images. By its essence the technique is invariant to rotation and translation. We first describe how the translational information content of an image may be defined, identified, assessed and analysed. The effect of the size is then assessed and quantified which permits us to propose a modified definition of translational information content invariant to scale. In order to demonstrate the feasibility of the developed techniques, a series of tests were carried out on rectangular and triangular shapes. The results illustrate the power of the translational information assessment techniques in analysing and describing 2-D images. The use of these techniques in shape description is also introduced.