Title :
Gradient methods for superresolution
Author :
Connolly, T.J. ; Lane, R.G.
Author_Institution :
Dept. of Electr. & Electron. Eng., Canterbury Univ., Christchurch, New Zealand
Abstract :
Conjugate gradient methods for superresolution are shown to accelerate convergence to the solution. The ill-posed nature of superresolution, combined with the fast convergence of the conjugate gradient algorithm, results in oscillatory artifacts or “null objects” which must be dealt with by regularization. We utilize a combination of Tikhonov-Miller regularization and positivity constraints as a means of regularizing the conjugate gradient algorithm
Keywords :
conjugate gradient methods; convergence of numerical methods; image resolution; Tikhonov-Miller regularization; conjugate gradient methods; fast convergence; image processing; null objects; oscillatory artifacts; positivity constraints; superresolution; Acceleration; Equations; Fourier transforms; Gradient methods; Image reconstruction; Image resolution; Image segmentation; Iterative methods; Large-scale systems; Spatial resolution;
Conference_Titel :
Image Processing, 1997. Proceedings., International Conference on
Conference_Location :
Santa Barbara, CA
Print_ISBN :
0-8186-8183-7
DOI :
10.1109/ICIP.1997.648116
