Optimal zone coding using the slant transform (Q1905173)

A special representative of the class of discrete orthogonal transforms with nonsinusoidal basis functions is the slant transform, which is distinguished by the presence of a slanted vector with linearly decreasing components in its basis. The slant transform of fourth and eighth orders was introduced in 1971 by \textit{H. Enomoto} and \textit{K. Shibata} [``Orthogonal transform coding system for television signals'', IEEE Trans. Electromagn. Compatibility EMC-13, 11-17 (1971)]\ especially for efficient representation of the video signal in line sections with smooth variation of brightness. It has been used for television image coding. \textit{W. K. Pratt}, \textit{W. H. Chen}, and \textit{L. R. Welch} [``Slant transform image coding'', IEEE Trans. Commun. COM-22, 1075-1093 (1974)]\ generalized the slant transform to vectors of any dimension \(N= 2^n\) and two-dimensional arrays, and derived posteriori estimates of reconstruction error with zonal iamge compression (the zones were chosen by trial and error) for various transforms. These estimates show that, for the same \(N\) and the same compression ratio \(\tau\), the slant transform is inferior to the Karhunen-Loève transform and superior to Walsh and Fourier transforms. In this paper, we derive a priori estimates of the reconstruction error for the slant transform in zone coding and suggest an optimal technique for zone selection.