1. Low-Complexity Lossless Compression of Hyperspectral Imagery via Linear Prediction By: Fei Nan & Hani Saad Presented to: Dr. Donald Adjeroh

2. Hyperspectral Image Compression 2 Index  Hyperspectral Images, what are they?  Remote Sensors and Low-complexity Image Compression  Linear Prediction (LP)  Spectral Oriented Least Squares (SLSQ)  LP Implementation  SLSQ Implementation  Experimental Results  Improvements  References

3. Hyperspectral Image Compression 3 Hyperspectral Images  High-definition electro-optic images  Used in surveillance, geology, environmental monitoring, and meteorology  224 contiguous bands  3 or more consecutive scenes

4. Hyperspectral Image Compression 4 Remote Sensors & Low-complexity Image Compression   Hyperspectral sensors measure hundreds of wavelengths   Airborne vs. Satellite Sensors   Why low-complexity compression?

5. Hyperspectral Image Compression 5 Linear Prediction (LP)  Spatial correlation  Spectral correlation  LP Interband linear prediction for interband codingInterband linear prediction for interband coding Standard median predicton for intraband codingStandard median predicton for intraband coding

6. Hyperspectral Image Compression 6 Linear Prediction cont’d  Standard median predicton Used for intraband codingUsed for intraband coding X i,j,k X i-1,j,k X i,j-1,k X i-1,j-1,k

7. Hyperspectral Image Compression 7 Linear Prediction cont’d  Interband linear prediction Used for interband codingUsed for interband coding

8. Hyperspectral Image Compression 8 Spectral Oriented Least Squares (SLSQ) Prediction defined in two different enumerations for pixel: 1.Intraband enumeration 2.Interband enumeration

9. Hyperspectral Image Compression 9 LP Implementation  The first 2 conds apply to Interband. 2 nd cond can be skip when T= œ, given T gives best performance.  The 3 rd cond applies to Intraband(IB).

10. Hyperspectral Image Compression 10 SLSQ Implementation The distance of Interband and intraband are defined. The Predictor Error Matrix C and Matrix X The simplified form when we assigned M=4 and N=1.

11. Hyperspectral Image Compression 11 Experimental Results

12. Hyperspectral Image Compression 12 Experimental Results cont’d 128x128x224 LPSLSQ Cuprite1.9182.425 Jasper1.8502.364 Low Altitude 1.7082.131 Lunar Lake 2.0652.390

13. Hyperspectral Image Compression 13 Improvements  Using M=5 vs. M=4  Keeping N=1  Future improvements can include look-ahead prediction SLSQ2SLSQ1 Cuprite2.4432.425 Jasper2.3582.364 Low Altitude 2.1292.131 Lunar Lake 2.4132.390 Average2.332.32

14. Hyperspectral Image Compression 14 References  Randall B. Smith, Ph.D., 17 September 2001. MicroImages, Inc. Introduction to Hyperspectral Imaging with TNTmips. www.microimages.com www.microimages.com  Peg Shippert, Ph.D., Earth Science Applications Specialist Research Systems, Inc. Introduction to Hyperspectral Image Analysis.  Suresh Subramanian,, Nahum Gat, Alan Ratcliff, Michael Eismann. Real-time Hyperspectral Data Compression Using Principal Components Transformation