Main Page   Class Hierarchy   Alphabetical List   Compound List   File List   Compound Members   Related Pages  

RAosCDDinpainting Class Reference

#include <RAosCDDinpainting.hpp>

Inheritance diagram for RAosCDDinpainting:

Inheritance graph
[legend]
Collaboration diagram for RAosCDDinpainting:

Collaboration graph
[legend]
List of all members.

Detailed Description

Float AOS Perona with strong edge detection.

Author:
Bernard De Cuyper
Version:
0.055
Date:
11-14/09/2002
 
Purpose:                CDD inpainting may be combined directly with restoration in a single AOS process. ;-)

                        This can be done if we know which is the inpainting area to fill.
                        Inpainting goes better when noise is reduced in the image. This mean that we 
                        can benefit from restoring at the same time the non-inpainting area, the known image!

                        By chance the analysis of the AOS restoration process, demonstrate that the conductivity
                  is easily separated from the whole, and is STRICTLY SPOKEN LOCAL!!!
                  
                        This mean:

                        1) In the restoration area, the conductivity is modulated by edge indicators, which tend to stop
                 restoration at object boundaries.
                        2) In the inpainting area, the conductivity is modulated by the absolute value of the curvature.
                 This allows the filling procedure to take care of the connectivity principle. :-) :-) :-)


                        Perona with strong edge detection

                        Fast Computation of Perona recursive flow.
                        Thomas LU model.
                        Semi-Implicit solver using AOS additive splitting

            (I - tau * A) * uNew=  uOld

                        Should be better than LOD: Rotation invariant.

                        LOD is sequential (handelling each direction x y z in sequence)

                        AOS is parallel (handelling each direction x y z at the same time)
                        AOS build an average operator

                        Both LOD and AOS are designed for large image restoration, O(N) in time ang space!!!

Papers:         "Recursivity and PDE's in image processing", 
                        L. Alvarez, R. Deriche and F Santana, Spain 1998.

                        "Efficient and Reliable Schemes for Nonlinear Diffusion Filtering", 
                        Joachim. Weickert & all, IEEE transactions on Image Processing, vol7, n3, March 1998.
                

@ Copyrights: Bernard De Cuyper & Eddy Fraiha 2002, Eggs & Pictures. MIT/Open BSD copyright model.


Public Methods

 RAosCDDinpainting (int iterations=8, double t=0.5, double deltaGradient=16.0, double asigma=5.0, bool iirFlag=false)
 RAosCDDinpainting (AbsRImageFlow *annihilator, int iterations=8, double t=0.5, double deltaGradient=16.0, double asigma=5.0, bool iirFlag=false)
virtual ~RAosCDDinpainting ()
virtual RImageselectedFilter (AnImage *mask, RImage *src, RImage *dest=0)
 Local filtering in a mask area.

virtual void report (FILE *file)

Protected Methods

virtual double g (int i, int j)

Private Attributes

AbsRImageFlowcurvatureLikeFlow
AnImageinpaintingArea
RImagef


Member Function Documentation

RImage * RAosCDDinpainting::selectedFilter AnImage   mask,
RImage   src,
RImage   dest = 0
[virtual]
 

Local filtering in a mask area.

Parameters:
mask  is AnImage* is a ByteImage
src  is RImage* is RImage source channel
dest  is RImage* is RImage result/placeholder
Returns :
RImage* as result,

Reimplemented from RSimpleAosOp.


The documentation for this class was generated from the following files:
SourceForge.net Logo
Restoreinpaint sourceforge project `C++/Java Image Processing, Restoration, Inpainting Project'.

Bernard De Cuyper: Open Project Leader: Concept, design and development.
Bernard De Cuyper & Eddy Fraiha 2002, 2003. Bernard De Cuyper 2004. Open and free, for friendly usage only.
Modifications on Belgium ground of this piece of artistic work, by governement institutions or companies, must be notified to Bernard De Cuyper.
bern_bdc@hotmail.com