<< Chapter < Page Chapter >> Page >
The total reconstruction error of transform coding with optimal bit allocation is compared to that of uniformly quantized PCM, and the ratio is found to depend on a type of spectral flatness measure.
  • With an orthogonal transform and the optimal bit allocation equation 8 from Optimal Bit Allocation (lower equation), the total reconstruction error equals
    σ r 2 | T C = 1 N = 0 N - 1 σ q 2 = γ y 2 - 2 R k = 0 N - 1 σ y k 2 1 / N .
    We can compare to uniformly quantized PCM, where σ r 2 | P C M = γ x σ x 2 2 - 2 R . Since an orthogonal transform implies
    σ x 2 = 1 N k = 0 N - 1 σ y k 2 ,
    we have the following gain over PCM:
    G TC = σ r 2 | P C M σ r 2 | T C = γ x γ y 1 N k = 0 N - 1 σ y k 2 k = 0 N - 1 σ y k 2 1 / N .
    Note that the gain is proportional to the ratio between arithmetic and geometric means of the transform coefficient variances.(Note similarities to the spectral flatness measure.) The factor γ y / γ x accounts for changes in distribution which affect uniform-quantizer efficiency.For example, if T caused uniformly distributed x to become Gaussian distributed y k , γ y / γ x would contribute a 7 dB loss in TC-to-PCM performance.If, on the other hand, x was Gaussian, then y k would also be Gaussian and γ y / γ x = 1 .

Get Jobilize Job Search Mobile App in your pocket Now!

Get it on Google Play Download on the App Store Now




Source:  OpenStax, An introduction to source-coding: quantization, dpcm, transform coding, and sub-band coding. OpenStax CNX. Sep 25, 2009 Download for free at http://cnx.org/content/col11121/1.2
Google Play and the Google Play logo are trademarks of Google Inc.

Notification Switch

Would you like to follow the 'An introduction to source-coding: quantization, dpcm, transform coding, and sub-band coding' conversation and receive update notifications?

Ask