AN ANALYSIS ON SYNTHETIC APERTURE RADAR DATA AND ENHANCEMENT OF RECONSTRUCTED IMAGES Cihn Erş e-mil: ers@eh.itu.edu.tr Istnul Technicl University, Fculty of Electricl nd Electronics Engineering, Deprtment of Electronics nd Communiction Engineering, 34469, Mslk, Istnul, Turkey Keywords: Dt nlysis, imge enhncement, histogrm expnsion, contrst correction ABSTRACT In this pper, properties of rw nd processed synthetic perture rdr ( SAR ) dt re discussed. Sttisticl nlyses re crried out on ERS-2 dt to understnd the nture of the dt of interest. Best pproch to chieve this gol is to nlyze the relted histogrms Histogrms of vrious smple dt re plotted nd checked whether they re similr to some known distriutions, prmeters needed to define these distriutions re determined. To improve visul qulity of the reconstructed imges histogrm expnsion, which is glol enhncement method, is used. I. INTRODUCTION Synthetic perture rdr techniques nd its pplictions were much emphsized in the field of remote sensing in recent yers. SAR sensors re eing developed s key mpping source dy y dy s their high qulity imging cpility is relized nd improved. While technologicl progresses re rpidly extended, oth irorne nd spceorne SAR systems demonstrte their fesiility nd more engineering projects in this re is developed. A SAR imging sensor synthesizes long ntenn y trnsmitting pulsed signls nd coherently dding the reflected signls successively to otin high resolution in zimuth direction. The resolution in rnge direction is chieved y trnsmitting very short pulses or wide ndwidth pulses ( chirp ). Thus, it provides informtion out the surfce y mesuring nd mpping the reflected energy in the microwve region. The wide vrieties of potentil pplictions of SAR imging cused the development of the SEASAT stellite. The lunch of the SEASAT stellite into the Erth orit y NASA in June 1978, provided n opportunity to investigte the utility of SAR for environmentl monitoring nd ocen oservtion from spce which ws impossile efore. During its mission period of three months, lrge mount of imgery ws otined nd it ws shown tht spceorne SAR ws n dvntgeous choice for glol monitoring of the Erth. To otin etter understnding of the spceorne SAR remote sensing, series of Shuttle Imging Rdr ( SIR ) flights, which include the 1981 SIR-A, 1984 SIR-B etc., ws performed. Recognizing the utilities of SAR, different countries developed their own spceorne SAR sensors to otin high resolution imges of the Erth for civilin pplictions. Russi lunched in 1991 ALMAZ-1, Jpn in 1992 JERS-1 nd Cnd in 1995 RADARSAT. Europen Union strted ERS-1 nd ERS-2 in 1991 nd 1995, respectively. In this study, ERS-2 rw nd processed dt for chosen prt of Istnul re nlyxed y histogrm nlysis. Vrious histogrms re plotted for the mentioned dt nd prmeters re defined. In imge enhncement steps, histogrm expnsion nd contrst modifiction re used. As result, visul ppernce of the imges re considerly improved. II. STATISTICAL ANALYSIS Histogrm nlysis is remrkle wy to nlyze imges sttisticlly. A histogrm is constructed y exmining the gry level vlue of ech pixel in the imge nd counting the numer of pixels displying ech of the possile vlues. On the histogrm plot, ech gry level vlue is represented y histogrm in whose height represents the numer of imge pixels displying tht vlue. Therefore the process of constructing the histogrm corresponds to the filling of histogrm ins. Histogrm of n imge gives n estimte of the proility of occurrence of gry levels. A plot of this function for ll gry levels provides glol description of the ppernce of the imge. For exmple, histogrm on which the gry levels re concentrted towrd the drk end of the gry scle rnge mens n imge with overll drk chrcteristics. Opposite is true s well. A histogrm which hs nrrow shpe indictes little dynmic rnge nd thus corresponds to n imge hving low contrst. As ll gry levels occur towrd the middle of the gry scle, the imge would pper gry. Although the properties discussed ove re glol descriptions tht sy nothing specific out imge content, the shpe of the histogrm of n imge does give us useful informtion out the possiility for contrst enhncement. As
we hve low contrst imges, contrst enhncement would e n importnt improvement fctor. To otin histogrms, we first normlize dt mtrix elements etween 0 255. Thus we hve 256 different gry level vlues. Then, histogrms re constructed y the properties discussed ove. Let us consider 1024x1024 ERS-2 rw dt nd the imge tht is produced vi nrrow focusing given in Figure 1. Figure 3. Histogrm of the processed imge Histogrms of the given rw nd processed imges cn e seen in Figure 2 nd Figure 3, respectively. It is relized tht processed imge is drk nd low contrst imge since its histogrm is mounted round drk gry levels nd hs nrrow pek. This histogrm hs men of 18. Distriution oserved in Figure 3 is quite similr to Ryleigh distriution which is defined s 2 x x ( ) p x =.exp 2, x 0 σ 2. σ 2 (1) Ryleigh distriution with the sme men ( 18 ) is given in Figure 4. (1) Figure 2.. Histogrm of the processed imge Figure 4. Ryleigh distriution
Normlized histogrm ccording to Ryleigh distriution cn e seen in Figure 5. Histogrm of the rw dt hs lso Ryleigh distriution which is the cse in Figure 2. Sme procedure cn e followed for rw dt s well. Liner mpping function which is used to expnd the exmple histogrm is; x x y = (2) x mx min.ymx xmin where x min nd x mx re the minimum nd mximum vlues of the input histogrm, respectively while y mx is the upper oundry for the output histogrm intervl [ 0, 255 ]. However, this mpping function is not efficient for nrrow-pek nd long- til histogrms since their gry level vlues rech to the minimum nd mximum of the full rnge. Thus, we cn define intervls for the input gry levels to eliminte the ones on histogrm tils. Firstly, we choose input histogrm intervl s [ 0, 150 ] y mpping the gry level vlues which re greter thn 150 to 150. As result, ll the vlues in dt mtrix re etween 0 nd 150. Then we pply the expnsion formul nd otin the output histogrm given in Figure 6. where the defined upper ound of the input histogrm is denoted y prmeter. Contrst improvement of the resulting imge is cler. Figure 5. Histogrm normlized y Ryleigh distriution III. IMAGE ENHANCEMENT The principle ojective of enhncement techniques is to process n imge so tht the result is more suitle thn the originl imge for specific ppliction. Imge enhncement is treted here s processing technique to increse the visul contrst of n imge in designted gry level rnge. Although the degree of enhncement my e sujective, procedures to perform given type of enhncement cn e directly relted to the desired purpose. The method descried here is histogrm trnsformtion, hence glol enhncement is presented. It genertes new output histogrm y modifying the shpe of the input histogrm ccording to specific mpping function tht is chosen to enhnce contrst in the rnge of interest. The result of this is tht the gry level vlues of the originl imge re modified to improve its ppernce or effectiveness for visul nlysis. The method pplied here is histogrm expnsion. It is the most strightforwrd trnsformtion for enhncement. This is liner trnsformtion tht entils n input intensity rnge to n output intensity rnge, which typiclly uses the full dynmic rnge. Rememer the imge in Figure 1. nd its histogrm. The histogrm hs nrrow pek, which mens the imge is low contrst imge. Hence, histogrm expnsion cn e pplied on this histogrm to use the full dynmic rnge ( tht is gry levels re etween 0 255 ). Figure 6. Improved imge () nd its histogrm vi histogrm expnsion () with prmeter = 150
As second choice, intervl for the input histogrm is determined s [ 0, 120 ]. Gry levels greter thn 120 re mpped to 120 in the dt mtrix. Appliction of the formul results in Figure 7. Figure 8. Improved imge () nd its histogrm vi histogrm expnsion () with prmeter = 80 Figure 7. Improved imge () nd its histogrm vi h= 120 Sme procedure is repeted for prmeter vlue of 80. Otined results re shown in Figure 8. Note the pek on the gry level 256, this effect occurs ecuse ll the vlues of 80 in the mtrix re mpped to 256. Also, s the chosen upper ound is lower, more significnt effect on the imge is oserved. Another enhncement pproch is to pply contrst correction to the imge efore ppliction of histogrm expnsion. Contrst correction is histogrm opertion which is used to enhnce the visul ppernce of n imge. Contrst modifiction of n imge is defined s ( g vg) vg si = m. i + (3) where s i is the i th grylevel vlue of the contrst enhnced imge. g i is the i th grylevel vlue of the originl imge, m is contrst fctor nd vg is the men vlue of the originl imge. Resulting imge nd histogrm for m=2.5 nd prmeter=150 is shown in Figure 9. Grylevel vlues less thn vg re mpped to zero since negtive vlues re experienced in this rnge. Figure 9. Improved imge () nd its histogrm vi histogrm expnsion () with m=2.5 nd prmeter = 150
IV. CONCLUSION A discussion on synthetic perture rdr dt is presented. Different spects of given ERS-2 dt re explined y histogrm sed methods. Proposed pproches re used for the nlysis of mentioned dt nd visul improvement of reconstructed imges. It is relized tht plotted histogrms re similr to some known distriutions which hve specific prmeters. Once the imges re formed, numer of post-processing steps re used to mximize the usefulness of the informtion in the rdr imgery. Visul enhncement on generted imges is chieved. REFERENCES 1. Elchi C., Bicknell T., Jordn R.L. nd Wu C., Spceorne Synthetic Aperture Imging Rdrs: Applictions, Techniques nd Technology, Proceedings of the IEEE, Vol.70, pp.1174 1209, 1982.. 2. Seul M., O Gormn L. nd Smmon M.J., Prcticl Algorithms for Imge Anlysis, Cmridge University Press, New York, 2000. 3. Myler H.R., The Pocket Hndook of Imge Processing Algorithms in C, Prentice Hll, New Jersey, 1993. 4. Gonzlez R.C. nd Woods R.E., Digitl Imge Processing, Addison Wesley, Cliforni, 1993. 5. Minhui Z., Advnce in Synthetic Aperture Rdr Activities, Microwve Conference 2000 Asi Pcific, pp.474 476, 2000. 6. Curlnder J.C., Utiliztion of Spceorne SAR Dt for Mpping, IEEE Trnsctions on Geoscience nd Remote Sensing, Vol.22, pp.106 112, 1984. 7. Muff D.G., Blke A.P., Horne A.M., Spceorne SAR Autofocus, Recent Developments in Rdr nd Sonr Imging Systems: Wht s Next?, IEE Colloquium, Vol.11, pp.1110 1111, 1995. 8. Benz U., Strodl K. nd Morier A., A Comprison of Severl Algorithms for SAR Rw Dt Compression, IEEE Trnsctions on Geoscience nd Remote Sensing, Vol.33, pp.1266 1268, 1995. 9. Bmler R. nd Runge H., PRF Amiguity Resolving y Wvelength Diversity, IEEE Trnsctions on Geoscience nd Remote Sensing, Vol.29,pp. 997 1001, 1991. 10. Li F.K. nd Johnson W.T.K., Amiguities in Spceorne Synthetic Aperture Rdr Dt, IEEE Trnsctions on Aerospce nd Electronic Systems, Vol.19, pp.389 395, 1983. 11. Moreir A., Suppressing the Azimuth Amiguities in Synthetic Aperture Rdr Imges, IEEE Trnsctions on Geoscience nd Remote Sensing, Vol.31, pp.885 888, 1993. 12. Mdsen S.N., Estimting the Doppler Centroid of SAR Dt, IEEE Trnsctions on Aerospce nd Electronic Systems, Vol.25, pp.134 139, 1989. 13. Tomiysu K., Tutoril Review of Synthetic Aperture Rdr with Applictions to Imging of the Ocen Surfce, Proceedings of the IEEE, Vol.66, pp.563 572, 1978.. 14. Berrdino P., Mrrzzo M., Migliccio M., Lnri R., ERS-1 / 2 nd IFSAR for Antrctic, IGARSS 01, Vol.6, pp. 2554 2556, 2001. 15. Frncheschetti G., Lnri R. nd Mrzouk E.S., Aerrtion free SAR Rw Dt Processing vi Trnsformed Grid Predeformtion, IGARSS 93, Vol.4, pp.1593 1595, 1993. 16. Frnceschetti G., Migliccio M., Riccio D.. Schirinzi G., SARAS: A Synthetic Aperture Rdr ( SAR ) Rw Signl Simultor, IEEE Trnsctions on Geoscience nd Remote Sensing, Vol.30, pp.110-113, 1992. 17. Goj, W.W., Synthetic Aperture Rdr nd Electronic Wrfre, Artech House, 1993.. 18. Mens,.High Resolution Rdr Imging, Artech House, 1981. 19. Curlnder, J.C. nd McDonough, R.N., Synthetic Aperture Rdr: Systems nd Signl Processing,Wiley&Sons, 1991. 20. Frnceschetti, G. nd Lnri R., Synthetic Aperture Rdr Processing, CRC Press, 1999.. 21. Morchin W., Rdr Engineer s Sourceook, Artech House, 1993..