WAVELET-BASED HYPERSPECTRAL IMAGE CODING USING ROBUST FIXED RATE TRELLIS-CODED QUANTIZATION

Publication Summary

A system is presented for compression of hyper- spectral imagery. Specifically, DPCM is used for spectral decorrelation, while a robust 2-D discrete wavelet coding scheme is used for spatial decorrela- tion. Trellis-coded quantization is used to encode the wavelet coefficients. Fixed-rate codebooks are designed using a modified version of the general- ized Lloyd algorithm. The system achieves a com- pression ratio of greater than 70: 1, with an average PSNR of the coded hyperspectral sequence exceed- ing40dR.

Country	United States
Language	English (United States)
Related Person(s)	(AUTHOR)  Glen P. Abousleman
Copyright	Motorola Inc. January 2000

---

Continue to Text Preview →

Page 1 of 4

M-LA Technical Developments

WAVELET-BASED HYPERSPECTRAL IMAGE CODING USING

ROBUST FIXED RATE TRELLIS-CODED QUANTIZATION

by Glen P. Abousleman

  A system is presented for compression of hyper- spectral imagery. Specifically, DPCM is used for spectral decorrelation, while a robust 2-D discrete wavelet coding scheme is used for spatial decorrela- tion. Trellis-coded quantization is used to encode the wavelet coefficients. Fixed-rate codebooks are designed using a modified version of the general- ized Lloyd algorithm. The system achieves a com- pression ratio of greater than 70: 1, with an average PSNR of the coded hyperspectral sequence exceed- ing40dR.

  A system was developed for compression of hyperspectral imagery. The algorithm is based on DPCM used in conjunction with the discrete wavelet transform (DWT). In this algorithm, DPCM is used for spectral decorrelation, and each "error image" is encoded using a robust 2-D DWT coder. The error image is decomposed into 22 subbands using a modified Mallat tree configuration. Each subband is all-pass filtered using a phase scrambling operation. All resulting sequences are quantized using fixed-rate trellis-coded quantization (FRTCQ). Codebooks are optimized for the Gaussian distribu- tion.

  Codebook design uses a modified version of the generalized Lloyd algorithm in a training-sequence- based iterative scheme. Rate allocation is per- formed in an optimal fashion by an iterative tech- nique, which uses the rate-distortion performance of the various trellis-based quantizers.

  DPCM is a simple and well-known method of achieving moderate compression of correlated sequences. Given a pixel, x1.,, the next pixel in the sequence, xi, is predicted. If xfii.i- is the predicted value of xi, then the difference, e, = xi - xili.i-, will, on average, be significantly smaller in magnitude than xi. Accordingly, fewer quantization bins, and thus, fewer bits are required to encode the error

sequence than would be required to encode the orig- inal sequence of pixels.

  It can be shown that for a nonzero-mean input sequence, the optimum (minimum MSE) first-order linear predictor is given by

q&l- = pi.1 + FL(l-P),

  where u and p are the mean and correlation coefficient of the sequence, respectively, and x~.,- is the predicted value of xi.

  Consider the encoder configuration shown in Figure 1. Here, the DPCM loop operates on entire images rather than on individual sequences. Given an image, x"., , the next image in the hyperspectral sequence can be estimated, and an "error image" can be formed from the difference e, = x, - xnln.i- The error image (at each instant in time) is spatially correlated and can be quantized using any image coding scheme. Note that the error image must be decoded within the encoder loop so that the quan- tized image, x,^, can be constructed and used to pre- dict the next image.

  The prediction error images have much lower energy than the original bands and can be subjected to very coarse quanti...

Download This Document →

Back to Overview