Transform Coding Using Discrete Tchebichef Polynomials

Moment functions based on Tchebichef polynomials have been used recently in pattern recognition applications. Such functions have robust feature representation capabilities needed for a recognition task. This paper explores the possibility of using orthonormal versions of Tchebichef polynomials for image compression. The mathematical framework for the definition of Tchebichef transforms is given, along with the various analytical properties, recurrence relations and transform equations. Initial experiments with gray level images have yielded promising results, with the Tchebichef transform giving a higher PSNR value compared to the cosine transform for certain image reconstructions.