Contribution to Data Compression

Abstract

In this thesis, two novel contributions for the enhancement of JPEG-based integertransforms are introduced. The first contribution consists of the development oftwo methods to generate integer Discrete Cosine Transform (DCT) approximations.The first proposed method is based on 16-points DCT and rounding-off operationsleads to two new low-complexity transforms. The second one consists of an introduction of null elements into a specified integer DCT leads to two other newlow-complexity, faster, and more efficient transforms. Therefore, the capability ofproposed transform matrices is improved in image compression applications.Thesecond contribution in this thesis consists of the development of a novel methodfor enhancement of integer-based DCT and integer-based DTT JPEG by generatinga multiplier-less approximate Joint Photographic Expert Group (JPEG) quantisationmatrix. The two proposed quantisation matrices have been successfully evaluatedagainst the conventional JPEG quantisation matrix for all different type test images.Experimental results show that the compression of greyscale images by the JPEGstandard based on the integer Discrete Tchebichev Transform (DTT) and combinedwith our proposed quantisation matrix leads to a performance improvement whencompared to the JPEG conventional luminance quantisation matrix while maintaining high visual quality of the reconstructed images