Near Optimal, Low Complexity Arithmetic Coding for Generalized Gaussian Sources
We propose a near optimal, asymmetrical, low complexity, block-based arithmetic coding algorithm for the prediction residuals produced by lossless or lossy audio compression algorithms. The analysis of real-world prediction residuals motivates the use of a modified two-sided continuous generalized Gaussian distribution, pdf(x)=c(p,s)*exp[-(|x|/s)^p], which is mapped to a discrete distribution. Closed form formulas and fast numerical estimates for the block parameters (p,s) are derived. Using a distribution property, precomputed probability tables are obtained for different values of p, independent of s. The compression performance is within 0.2% of a high complexity, symmetrical coder. For very low complexity codes, an alternate coding algorithm is also proposed, employing only bit output, with comparable performance.
Click to purchase paper or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $20 for non-members, $5 for AES members and is free for E-Library subscribers.