Lpfkeepswiththesame gainandothercopiesarerejected

Info iconThis preview shows page 1. Sign up to view the full content.

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

Unformatted text preview: h 20 5 9/11/2012 Interpolation ECE 6258 Fall 2012 • The goal of interpolation is to directly g[ n] f [ n] generate from . Interpolation consists of two steps: f [n ] g[ Ln ] F (u ) 1 | det L | |det L|1 Interpolation ECE 6258 Fall 2012 G (( Lt ) 1 (u li )) i 0 • This equation is a LPF when the copy of the spectra at the center. LPF keeps with the same gain and other copies are rejected. Ghassan AlRegib © Georgia Tech 21 Quantization ECE 6258 Fall 2012 • In the non‐saturation Mid‐tread Quantization regions, the max error is /2 • The number of levels can be carefully selected to avoid saturation. This require knowledge of and f max Ghassan AlRegib © Georgia Tech 22 Quantizer ECE 6258 Fall 2012 f min f max • If an image has values between and , and we can use an 8‐bit quantizer, then we are f min mapping to 00000000 and f max to 11111111, where is the quantizer step size f min Ghassan AlRegib © Georgia Tech 23 Ghassan AlRegib © Georgia Tech 24 6 9/11/2012 SNR ECE 6258 Fall 2012 • SNR (Signal to Noise Ratio): 2 f SNR 10 log10 2 dB e • If we have enough number of levels, the error has a uniform PDF: 1 e[m, n] [e[m, n]] 0 pe (e) 2 2 2 2 [e [m, n]] e 0 otherwise Ghassan AlRegib © Georgia Tech 2 12 25 SNR ECE 6258 Fall 2012 • Symmetric signal: f min f max 2 f max L L 2N # of bits • If the signal is uniform between f max and f max (2 f max ) 2 2 f max s2 N 12 2 2 N 2 /12 2 20 log 2 N 6.02 N dB SNR 10 log10 s2 10 log10 10 2 /12 e Ghassan AlRegib © Georgia Tech 26 PSNR ECE 6258 Fall 2012 • PSNR (Peak Signal to Noise Ratio): Peak Signal PSNR 10 log10 Peak Noise 2 Peak Signal 20 log10 Peak Noise 2 N 1 20 log10 ( ) /2 6.02 N dB Ghassan AlRegib © Georgia Tech 27 7...
View Full Document

Ask a homework question - tutors are online