Digital Image Compression using Hybrid Transform with Kekre Transform and Other Orthogonal Transforms

Abstract

This paper presents image compression technique using hybrid transform. Concept of hybrid wavelet transform can be extended to generate hybrid transform. In hybrid wavelet transform first few rows represent global features of an image and remaining rows represent local features of an image. In Hybrid wavelet matrix rows contributing to global characteristics can be varied. In the limiting case by taking kronecker product of to orthogonal component transforms, hybrid transform is generated where all rows of transform matrix represent global features and no local features are present. This hybrid transform matrix is then applied on color image. High frequency contents of transformed image are eliminated and only low frequency contents are retained to get compressed image. RMSE is calculated at different compression ratios to check the performance of hybrid transforms. Various orthogonal transforms like DCT,Walsh, Slant, Hartley, Real-DFT and DST are combined with Kekre transform to generate hybrid transforms. DKT-DCT gives better image quality and lower RMSE than other pairs formed with DKT. Component size 32-8 i.e.32x32(Kekre Transform) and 8x8 (DCT) gives best results than other possible size combinations like 8-32,16-16 and 64-4.