Reconstruction of random objects from noisy projections

✍ Scribed by M. Tasto

• Publisher: Elsevier Science
• Year: 1977
• Weight: 865 KB
• Volume: 6
• Category: Article
• ISSN: 0146-664X
• DOI: 10.1016/s0146-664x(77)80007-5

✦ Synopsis

A probabilistic object model for the performance analysis of vafimls algorithms of c;omput.erized transverse axial tomography (CTAT) is proposed. The model allows an tmalysis of the expected mem~ square reconstruction error as a htncA,ion of system paralnete, rs ~md obieet properties, namely, eorreh~tion. It is shown that the variance of ~]le i'eeo/lsbrtt(t{ioll otTer depends ett the IIHI{;llt~| correlations between setl~llnet' Oll~l~ii(;N attd on t,he e~orrelat.ioa between semmer outputs a,~d the random varlN~le at the l)uing t;o be reconstructed. Based on these results~ a rec~'cmstrue, tion algm'ithm in deduced which is optimM in the sense ef maximum likelihood and expected mean square reeonstru<;tiol~ error for both clean and noisy pm]ec;I.ions. The weights of the reeonstructio,t algorithm depend on the obje~t correlation, mt ~he noise, and on the number of projections used. An analytic expression for the reconstruetiml error iv also obtained, and some simple munerie~l extra,pies are presenLed.