| Summary: | 碩士 === 國立成功大學 === 應用數學研究所 === 82 === Image restoration is to improve a photographic image under some
criterions of goodness in reality. Restoration techniques are
oriented toward applying the inverse process of degraded model
in order to recover the original image from blur image.
Degraded model can be formulated as blur integral equation for
the purpose of the probelm. Now, we present a new choice of
solving the blur integral equation by finite element method
under the point spread function of out of focus blur. Thus we
focus on a more modern algebraic approach for the observed
large linear system with priori knowledge. On restoration
method, we derive a new reblur operator H^2 more stable than
previous H^TH from blur integral equation by finite element
directly. We develop an approximation algorithm of estimating
the condition of convergence of previous H^TH. Then two
iterative algorithms CGM( conjugate gradient method ) and PCGM(
preconditioned conjugate gradient method ) are applying to the
large linear system. In experiment, we try to test our
restoration method through several levels of additive noises.
At last, we do error analysis on H^2 and H^TH in view of norms
and then complete analysis of performance measure of
approximation algorithm.
|
|---|
