Introduction. A matrix method for constructing one-norm sine-cosine transforms of type II of order 4 has been developed, which has better efficiency compared to the known sine transform of type II. An integer one-norm sine-cosine transform of type II of order 4 has been proposed, and on its basis an integer one-norm simplified sine-cosine transform of order 8 of low computational complexity has been developed. Fast calculation algorithms for the proposed transforms have been considered. The computational complexity of the simplified sine-cosine transform of type II of order 8 is only 40 operations, which is three times less than the computational complexity of the known sine transform of type VII of order 8, and the compression ratio is 1.5-2.3% lower, as shown by the presented experimental results. The proposed integer one-norm simplified sine-cosine transform of order 8 can be used for image and signal analysis and coding tasks, in particular for separable adaptive transforms as an alternative to the sine transform of type VII for high-speed and extreme coding modes. Two simplified modes for adaptive separable transforms are proposed: mode A, which uses two combinations out of four total, namely 2D cosine and split cosine/sine; mode B, which uses three combinations out of four, excluding the sine/sine variation. As an alternative to the type VII sine transform in the separable transform scheme for adaptive application in high-speed modes, a modified type II sine transform is proposed, which consistently compresses 0.3 % better than the classic type II sine transform, and lags behind the type VII sine transform by 1.2–1.4 %. Keywords: discrete sine-cosine transform, integer sine-cosine transform, simplified sine-cosine transform, adaptive separable transforms, factorization, computational complexity, Hadamard transform, Haar transform, Haar-Hadamard transform, hybrid transforms.
Building similarity graph...
Analyzing shared references across papers
Loading...
Yaroslav Luts
V.M. Glushkov Institute of Cybernetics
Cybernetics and Computer Technologies
V.M. Glushkov Institute of Cybernetics
Building similarity graph...
Analyzing shared references across papers
Loading...
Yaroslav Luts (Fri,) studied this question.
synapsesocial.com/papers/69ccb75916edfba7beb893bd — DOI: https://doi.org/10.34229/2707-451x.26.1.6