Title :
Correcting general band-to-band misregistrations
Author_Institution :
Dept. of Electr. & Electron. Eng., Western Australia Univ., Nedlands, WA, Australia
Abstract :
Methods for estimating band-to-band misregistration are of two main types, cross-covariance or Fourier based. Here a general approach is outlined to optimise the use of either method for correcting complex misregistrations in a quick and straight forward manner. Consideration is given to possible feature matching problems associated with multi-resolution processing, and inherent problems with different types of imaging systems. The algorithm described is a simple divide and conquer system that uses predictions of local translations at continually increasing resolutions to form complex transformations. The method has been applied, and currently used to correct the complex misregistrations present in images captured using the DMSV Mkl camera system built by SpecTerra Systems Pty. Ltd
Keywords :
Fourier transforms; covariance analysis; divide and conquer methods; feature extraction; image matching; image registration; image resolution; prediction theory; spectral analysis; DMSV Mkl camera system; Fourier based method; SpecTerra Systems; algorithm; band to band misregistrations correction; complex transformations; crosscovariance method; divide and conquer system; feature matching problems; general approach; imaging systems; linear prediction; local translations prediction; multiresolution processing; mutlispectral images; Australia; Cameras; Data analysis; Image analysis; Information processing; Intelligent systems; Multispectral imaging; Optimization methods; Paper technology; Pixel;
Conference_Titel :
Image Processing, 1996. Proceedings., International Conference on
Print_ISBN :
0-7803-3259-8
DOI :
10.1109/ICIP.1996.560932
