source: trunk/src/civ_matlab.m @ 346

Last change on this file since 346 was 346, checked in by gostiaux, 12 years ago

Commited wrong file in previous commit; now OK

File size: 33.5 KB
Line 
1%'civ_matlab': Matlab version of the PIV programs CivX
2% --- call the sub-functions:
3%   civ: PIV function itself
4%   fix: removes false vectors after detection by various criteria
5%   patch: make interpolation-smoothing
6%------------------------------------------------------------------------
7% function [Data,errormsg,result_conv]= civ_uvmat(Param,ncfile)
8%
9%OUTPUT
10% Data=structure containing the PIV results and information on the processing parameters
11% errormsg=error message char string, default=''
12% resul_conv: image inter-correlation function for the last grid point (used for tests)
13%
14%INPUT:
15% Param: input images and processing parameters
16% ncfile: name of a netcdf file to be created for the result (extension .nc)
17%
18%AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
19%  Copyright  2011, LEGI / CNRS-UJF-INPG, joel.sommeria@legi.grenoble-inp.fr.
20%AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
21%     This is part of the toolbox UVMAT.
22%
23%     UVMAT is free software; you can redistribute it and/or modify
24%     it under the terms of the GNU General Public License as published by
25%     the Free Software Foundation; either version 2 of the License, or
26%     (at your option) any later version.
27%
28%     UVMAT is distributed in the hope that it will be useful,
29%     but WITHOUT ANY WARRANTY; without even the implied warranty of
30%     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
31%     GNU General Public License (open UVMAT/COPYING.txt) for more details.
32%AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
33
34function [Data,errormsg,result_conv]= civ_matlab(Param,ncfile)
35errormsg='';
36Data.ListGlobalAttribute={'Conventions','Program','CivStage'};
37Data.Conventions='uvmat/civdata';% states the conventions used for the description of field variables and attributes
38Data.Program='civ_matlab';
39Data.CivStage=0;%default
40ListVarCiv1={'Civ1_X','Civ1_Y','Civ1_U','Civ1_V','Civ1_C','Civ1_F'}; %variables to read
41ListVarFix1={'Civ1_X','Civ1_Y','Civ1_U','Civ1_V','Civ1_C','Civ1_F','Civ1_FF'};
42mask='';
43maskname='';%default
44check_civx=0;%default
45check_civ1=0;%default
46check_patch1=0;%default
47
48if ischar(Param)
49    Param=xml2struct(Param);
50end
51
52%% Civ1
53if isfield (Param,'Civ1')
54    check_civ1=1;% test for further use of civ1 results
55    % caluclate velocity data (y and v in indices, reverse to y component)
56    [xtable ytable utable vtable ctable F result_conv errormsg] = civ (Param.Civ1);
57
58   % to try the reverse_pair method, uncomment below
59%     [xtable1 ytable1 utable1 vtable1 ctable1 F1 result_conv1 errormsg1] = civ (Param.Civ1);
60%     Param.Civ1.reverse_pair=1;
61%     [xtable2 ytable2 utable2 vtable2 ctable2 F2 result_conv2 errormsg2] = civ (Param.Civ1);
62%     xtable=[xtable1; xtable2];
63%     ytable=[ytable1; ytable2];
64%     utable=[utable1; -utable2];
65%     vtable=[vtable1; -vtable2];
66%     ctable=[ctable1; ctable2];
67%     F=[F1; F2];
68%     result_conv=[result_conv1; result_conv2];
69%     errormsg=[errormsg1; errormsg2];
70   
71   
72    if ~isempty(errormsg)
73        return
74    end
75    list_param=(fieldnames(Param.Civ1))';
76    Civ1_param=list_param;%default
77    for ilist=1:length(list_param)
78        Civ1_param{ilist}=['Civ1_' list_param{ilist}];
79        %         eval(['Data.Civ1_' list_param{ilist} '=Param.Civ1.' list_param{ilist} ';'])
80        Data.(['Civ1_' list_param{ilist}])=Param.Civ1.(list_param{ilist});
81    end
82    Data.ListGlobalAttribute=[Data.ListGlobalAttribute Civ1_param];% {'Civ1_Time','Civ1_Dt'}];
83    %     if exist('ncfile','var')% TEST for use interactively with mouse_motion (no file created)
84    Data.ListVarName={'Civ1_X','Civ1_Y','Civ1_U','Civ1_V','Civ1_C','Civ1_F'};%  cell array containing the names of the fields to record
85    Data.VarDimName={'NbVec1','NbVec1','NbVec1','NbVec1','NbVec1','NbVec1'};
86    Data.VarAttribute{1}.Role='coord_x';
87    Data.VarAttribute{2}.Role='coord_y';
88    Data.VarAttribute{3}.Role='vector_x';
89    Data.VarAttribute{4}.Role='vector_y';
90    Data.VarAttribute{5}.Role='warnflag';
91    Data.Civ1_X=reshape(xtable,[],1);
92    Data.Civ1_Y=reshape(Param.Civ1.ImageHeight-ytable+1,[],1);
93    Data.Civ1_U=reshape(utable,[],1);
94    Data.Civ1_V=reshape(-vtable,[],1);
95    Data.Civ1_C=reshape(ctable,[],1);
96    Data.Civ1_F=reshape(F,[],1);
97    Data.CivStage=1;
98   
99else
100    if exist('ncfile','var')
101        CivFile=ncfile;
102    elseif isfield(Param.Patch1,'CivFile')
103        CivFile=Param.Patch1.CivFile;
104    end
105    Data=nc2struct(CivFile,'ListGlobalAttribute','absolut_time_T0') %look for the constant 'absolut_time_T0' to detect old civx data format
106    if isfield(Data,'Txt')
107        errormsg=Data.Txt;
108        return
109    end
110    if ~isempty(Data.absolut_time_T0')%read civx file
111        check_civx=1;% test for old civx data format
112        [Data,vardetect,ichoice]=nc2struct(CivFile);%read the variables in the netcdf file
113    else
114        if isfield(Param,'Fix1')
115            Data=nc2struct(CivFile,ListVarCiv1);%read civ1 data in the existing netcdf file
116        else
117            Data=nc2struct(CivFile,ListVarFix1);%read civ1 and fix1 data in the existing netcdf file
118        end
119    end
120end
121
122%% Fix1
123if isfield (Param,'Fix1')
124    ListFixParam=fieldnames(Param.Fix1);
125    for ilist=1:length(ListFixParam)
126        ParamName=ListFixParam{ilist};
127        ListName=['Fix1_' ParamName];
128        eval(['Data.ListGlobalAttribute=[Data.ListGlobalAttribute ''' ParamName '''];'])
129        eval(['Data.' ListName '=Param.Fix1.' ParamName ';'])
130    end
131    if check_civx
132        if ~isfield(Data,'fix')
133            Data.ListGlobalAttribute=[Data.ListGlobalAttribute 'fix'];
134            Data.fix=1;
135            Data.ListVarName=[Data.ListVarName {'vec_FixFlag'}];
136            Data.VarDimName=[Data.VarDimName {'nb_vectors'}];
137        end
138        Data.vec_FixFlag=fix(Param.Fix1,Data.vec_F,Data.vec_C,Data.vec_U,Data.vec_V,Data.vec_X,Data.vec_Y);
139    else
140        Data.ListVarName=[Data.ListVarName {'Civ1_FF'}];
141        Data.VarDimName=[Data.VarDimName {'NbVec1'}];
142        nbvar=length(Data.ListVarName);
143        Data.VarAttribute{nbvar}.Role='errorflag';   
144        Data.Civ1_FF=fix(Param.Fix1,Data.Civ1_F,Data.Civ1_C,Data.Civ1_U,Data.Civ1_V);
145        Data.CivStage=2;   
146    end
147end   
148%% Patch1
149if isfield (Param,'Patch1')
150    if check_civx
151        errormsg='Civ Matlab input needed for patch';
152        return
153    end
154    check_patch1=1;
155    Param.Patch1
156    Data.ListGlobalAttribute=[Data.ListGlobalAttribute {'Patch1_Rho','Patch1_Threshold','Patch1_SubDomain'}];
157    Data.Patch1_Rho=Param.Patch1.SmoothingParam;
158    Data.Patch1_Threshold=Param.Patch1.MaxDiff;
159    Data.Patch1_SubDomain=Param.Patch1.SubdomainSize;
160    Data.ListVarName=[Data.ListVarName {'Civ1_U_Diff','Civ1_V_Diff','Civ1_X_SubRange','Civ1_Y_SubRange','Civ1_NbSites','Civ1_X_tps','Civ1_Y_tps','Civ1_U_tps','Civ1_V_tps'}];
161    Data.VarDimName=[Data.VarDimName {'NbVec1','NbVec1',{'NbSubDomain1','Two'},{'NbSubDomain1','Two'},'NbSubDomain1',...
162             {'NbVec1Sub','NbSubDomain1'},{'NbVec1Sub','NbSubDomain1'},{'Nbtps1','NbSubDomain1'},{'Nbtps1','NbSubDomain1'}}];
163    nbvar=length(Data.ListVarName);
164    Data.VarAttribute{nbvar-1}.Role='vector_x';
165    Data.VarAttribute{nbvar}.Role='vector_y';
166    Data.Civ1_U_Diff=zeros(size(Data.Civ1_X));
167    Data.Civ1_V_Diff=zeros(size(Data.Civ1_X));
168    if isfield(Data,'Civ1_FF')
169        ind_good=find(Data.Civ1_FF==0);
170    else
171       
172        ind_good=1:numel(Data.Civ1_X);
173    end
174    [Data.Civ1_X_SubRange,Data.Civ1_Y_SubRange,Data.Civ1_NbSites,FFres,Ures, Vres,Data.Civ1_X_tps,Data.Civ1_Y_tps,Data.Civ1_U_tps,Data.Civ1_V_tps]=...
175                            patch(Data.Civ1_X(ind_good)',Data.Civ1_Y(ind_good)',Data.Civ1_U(ind_good)',Data.Civ1_V(ind_good)',Data.Patch1_Rho,Data.Patch1_Threshold,Data.Patch1_SubDomain);
176      Data.Civ1_U_Diff(ind_good)=Data.Civ1_U(ind_good)-Ures;
177      Data.Civ1_V_Diff(ind_good)=Data.Civ1_V(ind_good)-Vres;
178      Data.Civ1_FF(ind_good)=FFres;
179      Data.CivStage=3;                             
180end   
181
182%% Civ2
183if isfield (Param,'Civ2')
184    par_civ2=Param.Civ2;
185    if ~check_civ1 || ~strcmp(par_civ1.filename_ima_a,par_civ2.filename_ima_a)
186            par_civ2.ImageA=imread(par_civ2.filename_ima_a);%read first image if not already done for civ1
187    end
188    if ~check_civ1|| ~strcmp(par_civ1.filename_ima_b,par_civ2.filename_ima_b)
189            par_civ2.ImageB=imread(par_civ2.filename_ima_b);%read second image if not already done for civ1
190    end
191%     stepx=str2double(par_civ2.dx);
192%     stepy=str2double(par_civ2.dy);
193    ibx2=ceil(str2double(par_civ2.ibx)/2);
194    iby2=ceil(str2double(par_civ2.iby)/2);
195    isx2=ibx2+2;
196    isy2=iby2+2;
197
198
199%         Data.Civ1_U_Diff=zeros(size(Data.Civ1_X));
200%         Data.Civ1_V_Diff=zeros(size(Data.Civ1_X));
201%         if isfield(Data,'Civ1_FF')
202%             ind_good=find(Data.Civ1_FF==0);
203%         else
204%             ind_good=1:numel(Data.Civ1_X);
205%         end
206%             [Data.Civ1_X_SubRange,Data.Civ1_Y_SubRange,Data.Civ1_NbSites,FFres,Ures, Vres,Data.Civ1_X_tps,Data.Civ1_Y_tps,Data.Civ1_U_tps,Data.Civ1_V_tps]=...
207%                                 patch(Data.Civ1_X(ind_good)',Data.Civ1_Y(ind_good)',Data.Civ1_U(ind_good)',Data.Civ1_V(ind_good)',Data.Patch1_Rho,Data.Patch1_Threshold,Data.Patch1_SubDomain);
208%         end
209%     shiftx=str2num(par_civ1.shiftx);
210%     shifty=str2num(par_civ1.shifty);
211% TO GET shift from par_civ2.filename_nc1
212    % shiftx=velocity interpolated at position
213    miniy=max(1+isy2+shifty,1+iby2);
214    minix=max(1+isx2-shiftx,1+ibx2);
215    maxiy=min(size(par_civ2.ImageA,1)-isy2+shifty,size(par_civ2.ImageA,1)-iby2);
216    maxix=min(size(par_civ2.ImageA,2)-isx2-shiftx,size(par_civ2.ImageA,2)-ibx2);
217    [GridX,GridY]=meshgrid(minix:par_civ2.Dx:maxix,miniy:par_civ2.Dy:maxiy);
218    PointCoord(:,1)=reshape(GridX,[],1);
219    PointCoord(:,2)=reshape(GridY,[],1);
220
221    Guess = tps_eval(PointCoord,[Data.Civ1_X_tps Data.Civ1_Y_tps])
222    Shiftx=Guess*Data.Civ1_U_tps;
223    Shifty=Guess*Data.Civ1_V_tps;
224   
225    if ~isempty(par_civ2.maskname)&& ~strcmp(maskname,par_civ2.maskname)% mask exist, not already read in civ1
226        mask=imread(par_civ2.maskname);
227    end
228    % caluclate velocity data (y and v in indices, reverse to y component)
229    [xtable ytable utable vtable ctable F] = civ (par_civ2.ImageA,par_civ1.ImageB,ibx2,iby2,isx2,isy2,shiftx,-shifty,PointCoord,str2num(par_civ1.rho),mask);
230    list_param=(fieldnames(par_civ1))';
231    list_remove={'pxcmx','pxcmy','npx','npy','gridflag','maskflag','term_a','term_b','T0'};
232    index=zeros(size(list_param));
233    for ilist=1:length(list_remove)
234        index=strcmp(list_remove{ilist},list_param);
235        if ~isempty(find(index,1))
236            list_param(index)=[];
237        end
238    end
239    for ilist=1:length(list_param)
240        Civ1_param{ilist}=['Civ1_' list_param{ilist}];
241        eval(['Data.Civ1_' list_param{ilist} '=Param.Civ1.' list_param{ilist} ';'])
242    end
243    if isfield(Data,'Civ1_gridname') && strcmp(Data.Civ1_gridname(1:6),'noFile')
244        Data.Civ1_gridname='';
245    end
246    if isfield(Data,'Civ1_maskname') && strcmp(Data.Civ1_maskname(1:6),'noFile')
247        Data.Civ1_maskname='';
248    end
249    Data.ListGlobalAttribute=[Data.ListGlobalAttribute Civ1_param {'Civ1_Time','Civ1_Dt'}];
250    Data.Civ1_Time=str2double(par_civ1.T0);
251    Data.Civ1_Dt=str2double(par_civ1.Dt);
252    Data.ListVarName={'Civ1_X','Civ1_Y','Civ1_U','Civ1_V','Civ1_C','Civ1_F'};%  cell array containing the names of the fields to record
253    Data.VarDimName={'NbVec1','NbVec1','NbVec1','NbVec1','NbVec1','NbVec1'};
254    Data.VarAttribute{1}.Role='coord_x';
255    Data.VarAttribute{2}.Role='coord_y';
256    Data.VarAttribute{3}.Role='vector_x';
257    Data.VarAttribute{4}.Role='vector_y';
258    Data.VarAttribute{5}.Role='warnflag';
259    Data.Civ1_X=reshape(xtable,[],1);
260    Data.Civ1_Y=reshape(size(par_civ2.ImageA,1)-ytable+1,[],1);
261    Data.Civ1_U=reshape(utable,[],1);
262    Data.Civ1_V=reshape(-vtable,[],1);
263    Data.Civ1_C=reshape(ctable,[],1);
264    Data.Civ1_F=reshape(F,[],1);
265    Data.CivStage=Data.CivStage+1;
266end
267
268%% Fix2
269if isfield (Param,'Fix2')
270    ListFixParam=fieldnames(Param.Fix2);
271    for ilist=1:length(ListFixParam)
272        ParamName=ListFixParam{ilist};
273        ListName=['Fix1_' ParamName];
274        eval(['Data.ListGlobalAttribute=[Data.ListGlobalAttribute ''' ParamName '''];'])
275        eval(['Data.' ListName '=Param.Fix2.' ParamName ';'])
276    end
277    if check_civx
278        if ~isfield(Data,'fix2')
279            Data.ListGlobalAttribute=[Data.ListGlobalAttribute 'fix2'];
280            Data.fix2=1;
281            Data.ListVarName=[Data.ListVarName {'vec2_FixFlag'}];
282            Data.VarDimName=[Data.VarDimName {'nb_vectors2'}];
283        end
284        Data.vec_FixFlag=fix(Param.Fix2,Data.vec2_F,Data.vec2_C,Data.vec2_U,Data.vec2_V,Data.vec2_X,Data.vec2_Y);
285    else
286        Data.ListVarName=[Data.ListVarName {'Civ2_FF'}];
287        Data.VarDimName=[Data.VarDimName {'nbvec2'}];
288        nbvar=length(Data.ListVarName);
289        Data.VarAttribute{nbvar}.Role='errorflag';   
290        Data.Civ2_FF=fix(Param.Fix2,Data.Civ2_F,Data.Civ2_C,Data.Civ2_U,Data.Civ2_V);
291        Data.CivStage=5;   
292    end
293   
294end   
295
296%% Patch2
297if isfield (Param,'Patch2')
298    Data.ListGlobalAttribute=[Data.ListGlobalAttribute {'Patch2_Rho','Patch2_Threshold','Patch2_SubDomain'}];
299    Data.Patch2_Rho=str2double(Param.Patch2.Rho);
300    Data.Patch2_Threshold=str2double(Param.Patch2.Threshold);
301    Data.Patch2_SubDomain=str2double(Param.Patch2.SubDomain);
302    Data.ListVarName=[Data.ListVarName {'Civ2_U_Diff','Civ2_V_Diff','Civ2_X_SubRange','Civ2_Y_SubRange','Civ2_NbSites','Civ2_X_tps','Civ2_Y_tps','Civ2_U_tps','Civ2_V_tps'}];
303    Data.VarDimName=[Data.VarDimName {'NbVec2','NbVec2',{'NbSubDomain2','Two'},{'NbSubDomain2','Two'},'NbSubDomain2',...
304             {'NbVec2Sub','NbSubDomain2'},{'NbVec2Sub','NbSubDomain2'},{'Nbtps2','NbSubDomain2'},{'Nbtps2','NbSubDomain2'}}];
305    nbvar=length(Data.ListVarName);
306    Data.VarAttribute{nbvar-1}.Role='vector_x';
307    Data.VarAttribute{nbvar}.Role='vector_y';
308    Data.Civ2_U_Diff=zeros(size(Data.Civ2_X));
309    Data.Civ2_V_Diff=zeros(size(Data.Civ2_X));
310    if isfield(Data,'Civ2_FF')
311        ind_good=find(Data.Civ2_FF==0);
312    else
313        ind_good=1:numel(Data.Civ2_X);
314    end
315    [Data.Civ2_X_SubRange,Data.Civ2_Y_SubRange,Data.Civ2_NbSites,FFres,Ures, Vres,Data.Civ2_X_tps,Data.Civ2_Y_tps,Data.Civ2_U_tps,Data.Civ2_V_tps]=...
316                            patch(Data.Civ2_X(ind_good)',Data.Civ2_Y(ind_good)',Data.Civ2_U(ind_good)',Data.Civ2_V(ind_good)',Data.Patch2_Rho,Data.Patch2_Threshold,Data.Patch2_SubDomain);
317      Data.Civ2_U_Diff(ind_good)=Data.Civ2_U(ind_good)-Ures;
318      Data.Civ2_V_Diff(ind_good)=Data.Civ2_V(ind_good)-Vres;
319      Data.Civ2_FF(ind_good)=FFres;
320      Data.CivStage=6;                             
321end   
322
323%% write result in a netcdf file if requested
324if exist('ncfile','var')
325    errormsg=struct2nc(ncfile,Data);
326end
327
328% 'civ': function piv.m adapted from PIVlab http://pivlab.blogspot.com/
329%--------------------------------------------------------------------------
330% function [xtable ytable utable vtable typevector] = civ (image1,image2,ibx,iby step, subpixfinder, mask, roi)
331%
332% OUTPUT:
333% xtable: set of x coordinates
334% ytable: set of y coordiantes
335% utable: set of u displacements (along x)
336% vtable: set of v displacements (along y)
337% ctable: max image correlation for each vector
338% typevector: set of flags, =1 for good, =0 for NaN vectors
339%
340%INPUT:
341% image1:first image (matrix)
342% image2: second image (matrix)
343% ibx2,iby2: half size of the correlation box along x and y, in px (size=(2*iby2+1,2*ibx2+1)
344% isx2,isy2: half size of the search box along x and y, in px (size=(2*isy2+1,2*isx2+1)
345% shiftx, shifty: shift of the search box (in pixel index, yshift reversed)
346% step: mesh of the measurement points (in px)
347% subpixfinder=1 or 2 controls the curve fitting of the image correlation
348% mask: =[] for no mask
349% roi: 4 element vector defining a region of interest: x position, y position, width, height, (in image indices), for the whole image, roi=[];
350function [xtable ytable utable vtable ctable F result_conv errormsg] = civ (par_civ)
351%this funtion performs the DCC PIV analysis. Recent window-deformation
352%methods perform better and will maybe be implemented in the future.
353
354%% prepare grid
355ibx2=ceil(par_civ.Bx/2);
356iby2=ceil(par_civ.By/2);
357isx2=ceil(par_civ.Searchx/2);
358isy2=ceil(par_civ.Searchy/2);
359shiftx=par_civ.Shiftx;
360shifty=-par_civ.Shifty;% sign minus because image j index increases when y decreases
361if isfield(par_civ,'Grid')
362    if ischar(par_civ.Grid)
363        par_civ.Grid;
364        par_civ.Grid=dlmread(par_civ.Grid);
365        par_civ.Grid(1,:)=[];%the first line must be removed (heading in the grid file)
366    end
367else% automatic measurement grid
368    ibx2=ceil(par_civ.Bx/2);
369    iby2=ceil(par_civ.By/2);
370    isx2=ceil(par_civ.Searchx/2);
371    isy2=ceil(par_civ.Searchy/2);
372    shiftx=par_civ.Shiftx;
373    shifty=-par_civ.Shifty;
374    miniy=max(1+isy2+shifty,1+iby2);
375    minix=max(1+isx2-shiftx,1+ibx2);
376    maxiy=min(par_civ.ImageHeight-isy2+shifty,par_civ.ImageHeight-iby2);
377    maxix=min(par_civ.ImageWidth-isx2-shiftx,par_civ.ImageWidth-ibx2);
378    [GridX,GridY]=meshgrid(minix:par_civ.Dx:maxix,miniy:par_civ.Dy:maxiy);
379    par_civ.Grid(:,1)=reshape(GridX,[],1);
380    par_civ.Grid(:,2)=reshape(GridY,[],1);
381end
382
383%% Default output
384nbvec=size(par_civ.Grid,1);
385xtable=par_civ.Grid(:,1);
386ytable=par_civ.Grid(:,2);
387utable=zeros(nbvec,1);
388vtable=zeros(nbvec,1);
389ctable=zeros(nbvec,1);
390F=zeros(nbvec,1);
391result_conv=[];
392errormsg='';
393
394%% prepare mask
395if isfield(par_civ,'Mask') && ~isempty(par_civ.Mask)
396    if strcmp(par_civ.Mask,'all')
397        return    % get the grid only, no civ calculation
398    elseif ischar(par_civ.Mask)
399        par_civ.Mask=imread(par_civ.Mask);
400    end
401end
402check_MinIma=isfield(par_civ,'MinIma');% test for image luminosity threshold
403check_MaxIma=isfield(par_civ,'MaxIma') && ~isempty(par_civ.MaxIma);
404
405%% prepare images
406if isfield(par_civ,'reverse_pair')
407    if par_civ.reverse_pair
408        if ischar(par_civ.ImageB)
409            temp=par_civ.ImageA;
410            par_civ.ImageA=imread(par_civ.ImageB);
411        end
412        if ischar(temp)
413            par_civ.ImageB=imread(temp);
414        end
415    end
416else
417    if ischar(par_civ.ImageA)
418        par_civ.ImageA=imread(par_civ.ImageA);
419    end
420    if ischar(par_civ.ImageB)
421        par_civ.ImageB=imread(par_civ.ImageB);
422    end
423end
424
425[npy_ima npx_ima]=size(par_civ.ImageA);
426if ~isequal(size(par_civ.ImageB),[npy_ima npx_ima])
427    errormsg='image pair with unequal size';
428    return
429end
430par_civ.ImageA=double(par_civ.ImageA);
431par_civ.ImageB=double(par_civ.ImageB);
432
433
434%% Apply mask
435    % Convention for mask
436    % mask >200 : velocity calculated
437    %  200 >=mask>150;velocity not calculated, interpolation allowed (bad spots)
438    % 150>=mask >100: velocity not calculated, nor interpolated
439    %  100>=mask> 20: velocity not calculated, impermeable (no flux through mask boundaries) TO IMPLEMENT
440    %  20>=mask: velocity=0
441checkmask=0;
442if isfield(par_civ,'Mask') && ~isempty(par_civ.Mask)
443   checkmask=1;
444   if ~isequal(size(par_civ.Mask),[npy_ima npx_ima])
445        errormsg='mask must be an image with the same size as the images';
446        return
447   end
448  %  check_noflux=(par_civ.Mask<100) ;%TODO: to implement
449    check_undefined=(par_civ.Mask<200 & par_civ.Mask>=100 );
450    par_civ.ImageA(check_undefined)=min(min(par_civ.ImageA));% put image A to zero (i.e. the min image value) in the undefined  area
451    par_civ.ImageB(check_undefined)=min(min(par_civ.ImageB));% put image B to zero (i.e. the min image value) in the undefined  area
452end
453
454%% compute image correlations: MAINLOOP on velocity vectors
455corrmax=0;
456sum_square=1;% default
457% vector=[0 0];%default
458for ivec=1:nbvec
459    iref=par_civ.Grid(ivec,1);% xindex on the image A for the middle of the correlation box
460    jref=par_civ.Grid(ivec,2);% yindex on the image B for the middle of the correlation box
461    %     xtable(ivec)=iref;
462    %     ytable(ivec)=jref;%default
463    if ~(checkmask && par_civ.Mask(jref,iref)<=20) %velocity not set to zero by the black mask
464        if jref-iby2<1 || jref+iby2>par_civ.ImageHeight|| iref-ibx2<1 || iref+ibx2>par_civ.ImageWidth% we are outside the image
465            F(ivec)=3;
466        else
467            image1_crop=par_civ.ImageA(jref-iby2:jref+iby2,iref-ibx2:iref+ibx2);%extract a subimage (correlation box) from image A
468            image2_crop=par_civ.ImageB(jref+shifty-isy2:jref+shifty+isy2,iref+shiftx-isx2:iref+shiftx+isx2);%extract a larger subimage (search box) from image B
469            image1_mean=mean(mean(image1_crop));
470            image2_mean=mean(mean(image2_crop));
471            %threshold on image minimum
472            if check_MinIma && (image1_mean < par_civ.MinIma || image2_mean < par_civ.MinIma)
473                F(ivec)=3;
474            end
475            %threshold on image maximum
476            if check_MaxIma && (image1_mean > par_civ.MaxIma || image2_mean > par_civ.MaxIma)
477                F(ivec)=3;
478            end
479        end
480       
481        if F(ivec)~=3
482            image1_crop=image1_crop-image1_mean;%substract the mean
483            image2_crop=image2_crop-image2_mean;
484            sum_square=sum(sum(image1_crop.*image1_crop));
485            %reference: Oliver Pust, PIV: Direct Cross-Correlation
486            result_conv= conv2(image2_crop,flipdim(flipdim(image1_crop,2),1),'valid');
487            corrmax= max(max(result_conv));
488            result_conv=(result_conv/corrmax)*255; %normalize, peak=always 255
489            %Find the correlation max, at 255
490            [y,x] = find(result_conv==255,1);
491            if ~isempty(y) && ~isempty(x)
492                try
493                    if par_civ.Rho==1
494                        [vector,F(ivec)] = SUBPIXGAUSS (result_conv,x,y);
495                    elseif par_civ.Rho==2
496                        [vector,F(ivec)] = SUBPIX2DGAUSS (result_conv,x,y);
497                    end
498                    utable(ivec)=vector(1)+shiftx;
499                    vtable(ivec)=vector(2)+shifty;
500                    xtable(ivec)=iref+utable(ivec)/2;% convec flow (velocity taken at the point middle from imgae1 and 2)
501                    ytable(ivec)=jref+vtable(ivec)/2;
502                    iref=round(xtable(ivec));% image index for the middle of the vector
503                    jref=round(ytable(ivec));
504                    if checkmask && par_civ.Mask(jref,iref)<200 && par_civ.Mask(jref,iref)>=100
505                        utable(ivec)=0;
506                        vtable(ivec)=0;
507                        F(ivec)=3;
508                    end
509                    ctable(ivec)=corrmax/sum_square;% correlation value
510                catch ME
511                    %                     vector=[0 0]; %if something goes wrong with cross correlation.....
512                    F(ivec)=3;
513                end
514            else
515                F(ivec)=3;
516            end
517        end
518    end
519   
520    %Create the vector matrix x, y, u, v
521end
522result_conv=result_conv*corrmax/(255*sum_square);% keep the last correlation matrix for output
523
524%------------------------------------------------------------------------
525% --- Find the maximum of the correlation function after interpolation
526function [vector,F] = SUBPIXGAUSS (result_conv,x,y)
527%------------------------------------------------------------------------
528vector=[0 0]; %default
529F=0;
530[npy,npx]=size(result_conv);
531
532% if (x <= (size(result_conv,1)-1)) && (y <= (size(result_conv,1)-1)) && (x >= 1) && (y >= 1)
533    %the following 8 lines are copyright (c) 1998, Uri Shavit, Roi Gurka, Alex Liberzon, Technion ï¿œ Israel Institute of Technology
534    %http://urapiv.wordpress.com
535    peaky = y;
536    if y <= npy-1 && y >= 1
537        f0 = log(result_conv(y,x));
538        f1 = real(log(result_conv(y-1,x)));
539        f2 = real(log(result_conv(y+1,x)));
540        peaky = peaky+ (f1-f2)/(2*f1-4*f0+2*f2);
541    else
542        F=-2; % warning flag for vector truncated by the limited search box
543    end
544    peakx=x;
545    if x <= npx-1 && x >= 1
546        f0 = log(result_conv(y,x));
547        f1 = real(log(result_conv(y,x-1)));
548        f2 = real(log(result_conv(y,x+1)));
549        peakx = peakx+ (f1-f2)/(2*f1-4*f0+2*f2);
550    else
551        F=-2; % warning flag for vector truncated by the limited search box
552    end
553    vector=[peakx-floor(npx/2)-1 peaky-floor(npy/2)-1];
554
555%------------------------------------------------------------------------
556% --- Find the maximum of the correlation function after interpolation
557function [vector,F] = SUBPIX2DGAUSS (result_conv,x,y)
558%------------------------------------------------------------------------
559vector=[0 0]; %default
560F=-2;
561peaky=y;
562peakx=x;
563[npy,npx]=size(result_conv);
564if (x <= npx-1) && (y <= npy-1) && (x >= 1) && (y >= 1)
565    F=0;
566    for i=-1:1
567        for j=-1:1
568            %following 15 lines based on
569            %H. Nobach ï¿œ M. Honkanen (2005)
570            %Two-dimensional Gaussian regression for sub-pixel displacement
571            %estimation in particle image velocimetry or particle position
572            %estimation in particle tracking velocimetry
573            %Experiments in Fluids (2005) 38: 511ï¿œ515
574            c10(j+2,i+2)=i*log(result_conv(y+j, x+i));
575            c01(j+2,i+2)=j*log(result_conv(y+j, x+i));
576            c11(j+2,i+2)=i*j*log(result_conv(y+j, x+i));
577            c20(j+2,i+2)=(3*i^2-2)*log(result_conv(y+j, x+i));
578            c02(j+2,i+2)=(3*j^2-2)*log(result_conv(y+j, x+i));
579        end
580    end
581    c10=(1/6)*sum(sum(c10));
582    c01=(1/6)*sum(sum(c01));
583    c11=(1/4)*sum(sum(c11));
584    c20=(1/6)*sum(sum(c20));
585    c02=(1/6)*sum(sum(c02));
586    deltax=(c11*c01-2*c10*c02)/(4*c20*c02-c11^2);
587    deltay=(c11*c10-2*c01*c20)/(4*c20*c02-c11^2);
588    if abs(deltax)<1
589        peakx=x+deltax;
590    end
591    if abs(deltay)<1
592        peaky=y+deltay;
593    end
594end
595vector=[peakx-floor(npx/2)-1 peaky-floor(npy/2)-1];
596
597%'RUN_FIX': function for fixing velocity fields:
598%-----------------------------------------------
599% RUN_FIX(filename,field,flagindex,thresh_vecC,thresh_vel,iter,flag_mask,maskname,fileref,fieldref)
600%
601%filename: name of the netcdf file (used as input and output)
602%field: structure specifying the names of the fields to fix (depending on civ1 or civ2)
603    %.vel_type='civ1' or 'civ2';
604    %.nb=name of the dimension common to the field to fix ('nb_vectors' for civ1);
605    %.fixflag=name of fix flag variable ('vec_FixFlag' for civ1)
606%flagindex: flag specifying which values of vec_f are removed:
607        % if flagindex(1)=1: vec_f=-2 vectors are removed
608        % if flagindex(2)=1: vec_f=3 vectors are removed
609        % if flagindex(3)=1: vec_f=2 vectors are removed (if iter=1) or vec_f=4 vectors are removed (if iter=2)
610%iter=1 for civ1 fields and iter=2 for civ2 fields
611%thresh_vecC: threshold in the image correlation vec_C
612%flag_mask: =1 mask used to remove vectors (0 else)
613%maskname: name of the mask image file for fix
614%thresh_vel: threshold on velocity, or on the difference with the reference file fileref if exists
615%inf_sup=1: remove values smaller than threshold thresh_vel, =2, larger than threshold
616%fileref: .nc file name for a reference velocity (='': refrence 0 used)
617%fieldref: 'civ1','filter1'...feld used in fileref
618
619function FF=fix(Param,F,C,U,V,X,Y)
620FF=zeros(size(F));%default
621Param
622
623%criterium on warn flags
624FlagName={'CheckFmin2','CheckF2','CheckF3','CheckF4'};
625FlagVal=[-2 2 3 4];
626for iflag=1:numel(FlagName)
627    if isfield(Param,FlagName{iflag}) && Param.(FlagName{iflag})
628        FF=(FF==1| F==FlagVal(iflag));
629    end
630end
631%criterium on correlation values
632if isfield (Param,'MinCorr')
633    FF=FF==1 | C<Param.MinCorr;
634end
635if (isfield(Param,'MinVel')&&~isempty(Param.MinVel))||(isfield (Param,'MaxVel')&&~isempty(Param.MaxVel))
636    Umod= U.*U+V.*V;
637    if isfield (Param,'MinVel')&&~isempty(Param.MinVel)
638        FF=FF==1 | Umod<(Param.MinVel*Param.MinVel);
639    end
640    if isfield (Param,'MaxVel')&&~isempty(Param.MaxVel)
641        FF=FF==1 | Umod>(Param.MaxVel*Param.MaxVel);
642    end
643end
644return
645
646
647FF=double(FF);
648
649
650
651%------------------------------------------------------------------------
652% patch function
653% OUTPUT:
654% SubRangx,SubRangy(NbSubdomain,2): range (min, max) of the coordiantes x and y respectively, for each subdomain
655% nbpoints(NbSubdomain): number of source points for each subdomain
656% FF: false flags
657% U_smooth, V_smooth: filtered velocity components at the positions of the initial data
658% X_tps,Y_tps,U_tps,V_tps: positions and weight of the tps for each subdomain
659%
660% INPUT:
661% X, Y: set of coordinates of the initial data
662% U,V: set of velocity components of the initial data
663% Rho: smoothing parameter
664% Threshold: max diff accepted between smoothed and initial data
665% Subdomain: estimated number of data points in each subdomain
666
667function [SubRangx,SubRangy,nbpoints,FF,U_smooth,V_smooth,X_tps,Y_tps,U_tps,V_tps] =patch(X,Y,U,V,Rho,Threshold,SubDomain)
668%subdomain decomposition
669warning off
670U=reshape(U,[],1);
671V=reshape(V,[],1);
672X=reshape(X,[],1);
673Y=reshape(Y,[],1);
674nbvec=numel(X);
675NbSubDomain=ceil(nbvec/SubDomain);
676MinX=min(X);
677MinY=min(Y);
678MaxX=max(X);
679MaxY=max(Y);
680RangX=MaxX-MinX;
681RangY=MaxY-MinY;
682AspectRatio=RangY/RangX;
683NbSubDomainX=max(floor(sqrt(NbSubDomain/AspectRatio)),1);
684NbSubDomainY=max(floor(sqrt(NbSubDomain*AspectRatio)),1);
685NbSubDomain=NbSubDomainX*NbSubDomainY;
686SizX=RangX/NbSubDomainX;%width of subdomains
687SizY=RangY/NbSubDomainY;%height of subdomains
688CentreX=linspace(MinX+SizX/2,MaxX-SizX/2,NbSubDomainX);
689CentreY=linspace(MinY+SizY/2,MaxY-SizY/2,NbSubDomainY);
690[CentreX,CentreY]=meshgrid(CentreX,CentreY);
691CentreY=reshape(CentreY,1,[]);% Y positions of subdomain centres
692CentreX=reshape(CentreX,1,[]);% X positions of subdomain centres
693rho=SizX*SizY*Rho/1000000;%optimum rho increase as the area of the subdomain (division by 10^6 to reach good values with the default GUI input)
694U_tps_sub=zeros(length(X),NbSubDomain);%default spline
695V_tps_sub=zeros(length(X),NbSubDomain);%default spline
696U_smooth=zeros(length(X),1);
697V_smooth=zeros(length(X),1);
698
699nb_select=zeros(length(X),1);
700FF=zeros(length(X),1);
701check_empty=zeros(1,NbSubDomain);
702SubRangx=zeros(NbSubDomain,2);%initialise the positions of subdomains
703SubRangy=zeros(NbSubDomain,2);
704for isub=1:NbSubDomain
705    SubRangx(isub,:)=[CentreX(isub)-0.55*SizX CentreX(isub)+0.55*SizX];
706    SubRangy(isub,:)=[CentreY(isub)-0.55*SizY CentreY(isub)+0.55*SizY];
707    ind_sel_previous=[];
708    ind_sel=0;
709    while numel(ind_sel)>numel(ind_sel_previous) %increase the subdomain during four iterations at most
710        ind_sel_previous=ind_sel;
711        ind_sel=find(X>=SubRangx(isub,1) & X<=SubRangx(isub,2) & Y>=SubRangy(isub,1) & Y<=SubRangy(isub,2));
712        % if no vector in the subdomain, skip the subdomain
713        if isempty(ind_sel)
714            check_empty(isub)=1;   
715            U_tps(1,isub)=0;%define U_tps and V_tps by default
716            V_tps(1,isub)=0;
717            break
718            % if too few selected vectors, increase the subrange for next iteration
719        elseif numel(ind_sel)<SubDomain/4 && ~isequal( ind_sel,ind_sel_previous);
720            SubRangx(isub,1)=SubRangx(isub,1)-SizX/4;
721            SubRangx(isub,2)=SubRangx(isub,2)+SizX/4;
722            SubRangy(isub,1)=SubRangy(isub,1)-SizY/4;
723            SubRangy(isub,2)=SubRangy(isub,2)+SizY/4;
724        else
725            [U_smooth_sub,U_tps_sub]=tps_coeff(X(ind_sel),Y(ind_sel),U(ind_sel),rho);
726            [V_smooth_sub,V_tps_sub]=tps_coeff(X(ind_sel),Y(ind_sel),V(ind_sel),rho);
727            UDiff=U_smooth_sub-U(ind_sel);
728            VDiff=V_smooth_sub-V(ind_sel);
729            NormDiff=UDiff.*UDiff+VDiff.*VDiff;
730            FF(ind_sel)=20*(NormDiff>Threshold);%put FF value to 20 to identify the criterium of elimmination
731            ind_ind_sel=find(FF(ind_sel)==0); % select the indices of ind_sel corresponding to the remaining vectors
732            % no value exceeds threshold, the result is recorded
733            if isequal(numel(ind_ind_sel),numel(ind_sel))
734                U_smooth(ind_sel)=U_smooth(ind_sel)+U_smooth_sub;
735                V_smooth(ind_sel)=V_smooth(ind_sel)+V_smooth_sub;
736                nbpoints(isub)=numel(ind_sel);
737                X_tps(1:nbpoints(isub),isub)=X(ind_sel);
738                Y_tps(1:nbpoints(isub),isub)=Y(ind_sel);
739                U_tps(1:nbpoints(isub)+3,isub)=U_tps_sub;
740                V_tps(1:nbpoints(isub)+3,isub)=V_tps_sub;         
741                nb_select(ind_sel)=nb_select(ind_sel)+1;
742                 display('good')
743                break
744                % too few selected vectors, increase the subrange for next iteration
745            elseif numel(ind_ind_sel)<SubDomain/4 && ~isequal( ind_sel,ind_sel_previous);
746                SubRangx(isub,1)=SubRangx(isub,1)-SizX/4;
747                SubRangx(isub,2)=SubRangx(isub,2)+SizX/4;
748                SubRangy(isub,1)=SubRangy(isub,1)-SizY/4;
749                SubRangy(isub,2)=SubRangy(isub,2)+SizY/4;
750%                 display('fewsmooth')
751                % interpolation-smoothing is done again with the selected vectors
752            else
753                [U_smooth_sub,U_tps_sub]=tps_coeff(X(ind_sel(ind_ind_sel)),Y(ind_sel(ind_ind_sel)),U(ind_sel(ind_ind_sel)),rho);
754                [V_smooth_sub,V_tps_sub]=tps_coeff(X(ind_sel(ind_ind_sel)),Y(ind_sel(ind_ind_sel)),V(ind_sel(ind_ind_sel)),rho);
755                U_smooth(ind_sel(ind_ind_sel))=U_smooth(ind_sel(ind_ind_sel))+U_smooth_sub;
756                V_smooth(ind_sel(ind_ind_sel))=V_smooth(ind_sel(ind_ind_sel))+V_smooth_sub;
757                nbpoints(isub)=numel(ind_ind_sel);
758                X_tps(1:nbpoints(isub),isub)=X(ind_sel(ind_ind_sel));
759                Y_tps(1:nbpoints(isub),isub)=Y(ind_sel(ind_ind_sel));
760                U_tps(1:nbpoints(isub)+3,isub)=U_tps_sub;
761                V_tps(1:nbpoints(isub)+3,isub)=V_tps_sub;
762                nb_select(ind_sel(ind_ind_sel))=nb_select(ind_sel(ind_ind_sel))+1;
763                display('good2')
764                break
765            end
766        end
767    end
768end
769ind_empty=find(check_empty);
770%remove empty subdomains
771if ~isempty(ind_empty)
772    SubRangx(ind_empty,:)=[];
773    SubRangy(ind_empty,:)=[];
774    X_tps(:,ind_empty)=[];
775    Y_tps(:,ind_empty)=[];
776    U_tps(:,ind_empty)=[];
777    V_tps(:,ind_empty)=[];
778end
779nb_select(nb_select==0)=1;%ones(size(find(nb_select==0)));
780U_smooth=U_smooth./nb_select;
781V_smooth=V_smooth./nb_select;
782
783
784
785
786
Note: See TracBrowser for help on using the repository browser.