eee508_ScalarQuantization

# eee508_ScalarQuantization - Qua t at o Quantization...

This preview shows pages 1–8. Sign up to view the full content.

Quantization Quantization: main component in source coders ¾ Problem: Original image is often very large ± Reduce the data 9 Remove redundancy ± no loss ± info content preserved Prediction (DPCM, ADPCM) Entropy coding Transform coding (e.g., transforms used to decorrelate data) subband coding 9 Remove unnecessary, “least important” info ± loss scalar quantization vector quantization EEE 508 - Lecture 7

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Scalar quantization Example : Uniform and non-uniform scalar quantization x Uniform Uniform quantization ¾ Better performance (in terms of quantization distribution) can be t 0 achieved by designing quantizer that match the pdf of the signal x EEE 508 - Lecture 7
Scalar quantization Example : Uniform and non-uniform scalar quantization Non-uniform x [ n ] Non uniform quantization ¾ Might be better to have lots of steps at low levels and a few at high n 0 levels ¾ Look at pdf of signal or image and match the quantizer to it EEE 508 - Lecture 7

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Scalar quantization General Approach: ¾ Take histogram of image and normalize it (so that area = 1) to obtain a pdf. Usually, it takes a form close to a Gaussian or a Laplacian distribution ¾ Then design quantizer (choose { V j } and { r j } ) to match pdf and minimize distortion ¾ To solve the obtained resulting equations is not impossible but they are non linear So usually the equations have been solved they are non-linear. So, usually, the equations have been solved numerically for different distributions and number of quantization levels and the parameters of optimal quantizers (Lloyd-Max) are tabulated. EEE 508 - Lecture 7
Scalar quantization Simple example G Q ( x ) = x Two level quantizer p ( x ) x input with pdf x - x ¾ What is the optimal reconstruction level ? ± ˆ ²³ ² ³ ´ f f ± ± dx x p x x e E E x x e e Q 2 2 2 ˆ 2 V EEE 508 - Lecture 7

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Scalar quantization ¾ Use symmetry of the problem: ± ² ± ² ½ ­ ³ x p x p ±² ± ² ¿ ¾ ¯ ® ³ ´ ³ ³ 2 2 )) ( ( )) ( ( x e x e x Q x Q f f 2 2 2 ˆ 2 d p d p E G ± ² ± ² ± ² ± ² µ µ ³ ³ 0 0 2 dx x x dx x x x Q ± ² ± ² ± ² 0 2 4 2 0 Take 2 2 » º « ª ³ ´ f f f dx x p dx x xp dx x p x d dE 0 0 0 ¼ ¬ µ µ µ d d ±² ±² 0 4 4 0 0 ³ ´ µ µ f f dx x xp dx x p 2 1 ±² ±³ ´ µ µ f f ´ ´ 2 4 2 dx x xp dx x xp EEE 508 - Lecture 7 0 0
Scalar quantization ± Apply same idea for the more general M-level case Note: In previous example, the quantization levels are fixed and the optimzation is performed with respect to the reconstruction levels assuming a uniform-quantizer structure. More generally, we would want to optimize with respect to both quantization levels { V j } and reconstruction levels { r j } ± Lloyd-Max Quantizer optimal in the mean square sense (mse distortion Quantizer optimal in the mean square sense (mse distortion minimized).

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

### Page1 / 28

eee508_ScalarQuantization - Qua t at o Quantization...

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online