[GRASS5] r.wavelet

[Andrea Antonello]

Alright, I did some check and will answer and correct myself.

> Yes, I simply included the r.wavelet folder in my raster folder and
> everything works.

Not true, you will need a file makefile with the following in it:

include head.mk

PGM1=r.biowave.dec
PGM2=r.owave.dec
PGM3=r.biowave.rec
PGM4=r.owave.rec
OBJ1=$(OBJARCH)/biowave2.o $(OBJARCH)/biofunctions.o $(OBJARCH)/sconvolve.o
OBJ2=$(OBJARCH)/owave2.o $(OBJARCH)/biofunctions.o $(OBJARCH)/sconvolve.o
OBJ3=$(OBJARCH)/ibiowave2.o $(OBJARCH)/biofunctions.o $(OBJARCH)/sconvolve.o
OBJ4=$(OBJARCH)/iowave2.o $(OBJARCH)/biofunctions.o $(OBJARCH)/sconvolve.o
all: $(BIN)/$(PGM1) $(BIN)/$(PGM2) $(BIN)/$(PGM3) $(BIN)/$(PGM4)

$(BIN)/$(PGM1):$(OBJ1) $(GISLIB)
 $(CC) $(LDFLAGS) -o $@ $(OBJ1) $(GISLIB) $(MATHLIB) $(XDRLIB) $(RASTERLIB)
$(BIN)/$(PGM2):$(OBJ2) $(GISLIB)
 $(CC) $(LDFLAGS) -o $@ $(OBJ2) $(GISLIB) $(MATHLIB) $(XDRLIB) $(RASTERLIB)
$(BIN)/$(PGM3):$(OBJ3) $(GISLIB)
 $(CC) $(LDFLAGS) -o $@ $(OBJ3) $(GISLIB) $(MATHLIB) $(XDRLIB) $(RASTERLIB)
$(BIN)/$(PGM4):$(OBJ4) $(GISLIB)
 $(CC) $(LDFLAGS) -o $@ $(OBJ4) $(GISLIB) $(MATHLIB) $(XDRLIB) $(RASTERLIB)
$(OBJ1):biowave2.h
$(OBJ2):biowave2.h
$(OBJ3):biowave2.h
$(OBJ4):biowave2.h
$(GISLIB):

include tail.mk


> Some lines of background (very fast, give me the chance :o)): the wavelet
> filtering process creates from one map four maps of half of the resolution.
> So every processing level takes map
>
> 1
>
> and creates four maps:
>
> 2_11, 2_12, 2_21, 2_22
>
> that have obviously same boundaries but different (half) resolution.
> Next level takes the map of the lower resolution map
>
> 2_22   (the others contain the difference values in x, y and xy)

This was wrong, the lower resolution one is in the *_11, i.e. 2_11. *_22 is 
the xy diagonal direction.


Alright, now the solution comes, even if it's a workaround version. Please 
take into account that when these modules were written, I was at my very 
first programming and grassprogramming experiences.

Btw, the fact is that the modules work only for floating point FCELLs, i.e. 
not for integers and not for doubles. This is for sure due to some casting 
problem I could not deal with at the time (I feel ashamed...).

The working result can be seen on 6 levels of decomposition the concave-convex 
zones of an island (twice you can guess which one :)):

http://www.hydrologis.com/testfiles/waveletsscreens/images.html

Next question will be: when will you fix this?
I'm very busy at the moment and will not be able to fix that soon (did you 
ever go through the code of a beginning programmer?). So the only way now is 
to deal with floating point maps. Sorry for that.

Cheers,
Andrea












>
> So the boundary are the same, but when I visualize, the result is the
> following:
>
> http://www.hydrologis.com/testfiles/wavelet_pbl.png
>
>
> The bigger map is the third level of wavelet decomposition of the smaller
> map. The boundaries are the same, so I can't understand why the lower
> resolution map gets stretched.
>
> In the last years did something change in the raster apis, which I missed?
>
> Sajith, can you try and tell me if you get the same results?
>
> Has anyone some hint? I'm not really happy to go through some code I didn't
> open for years if it is not really necessary.
>
> > Also why r.wavelet is not part of grass6 distribution?
>
> I guess because they seem not to work at the time :)
> Honestly you are the second person that ask for multiresolution analysis in
> 3 years, so I think that is why they are in the add-ons section.
>
> > Does it lacks something, like documentation?
>
> No, the manual pages were made properly and exaustive.
>
>
> Hope this can be solved,
> Cheers,
> Andrea
>
> > On 7/3/05, Andrea Antonello <moovida at katamail.com> wrote:
> > > Hi Sajith,
> > >
> > > >        I am planing to do some image processing based on wavelet.
> > > > I checked the r.wavelet module in the grass addon section, but it is
> > > > made for grass 4(?). I wish to upgrade it to the present grass
> > > > version.
> > >
> > > the modules were made for the grass5 version. I got in touch only once
> > > again with the wavelets modules in grass6, where they gave some strange
> > > output, but at that time I had no second free to understand what it
> > > was. Since the raster api didn't change (someone correct me if it's not
> > > true), those modules should work out of the box.
> > >
> > > > I am not a grass developer now, and I dont know the changes in grass
> > > > libraries between version 4 and 6.
> > > > Can some one guide me on how to proceed?
> > >
> > > What do you mean? Compilation into the release or usage?
> > > The man pages in the package should fullfill the second.
> > > However for a first quick use:
> > > r.biowave.dec input=<mapname> output=<outputname> filter1=sp02.ir
> > > filter2=sl05.ir filterpath=<folder where the filters (sp02.ir and
> > > sl05.ir) are with / at the end>  NumRec=3
> > >
> > > This will perform a 3 step wavelets decomposition.
> > > The maps will be created and the names are selfexplanatory.
> > >
> > > If I remeber correct, the starting map has to be floating point, else
> > > there will be strange results.
> > >
> > > Cheers,
> > > Andrea
> > >
> > >
> > >
> > > --
> > > _______________________________________________________________________
> > >__ ___ HydroloGIS - Environmental Open Source Solutions
> > > www.hydrologis.com
> > >
> > > Andrea Antonello
> > > Environmental Engineer
> > > mobile:  +393288497722
> > >
> > > "Let it be as much a great honour to take as to give learning,
> > > if you want to be called wise."
> > > Skuggsja' - The King's mirror - 1240 Reykjavik
> > > _______________________________________________________________________
> > >__ ___

-- 
____________________________________________________________________________
HydroloGIS - Environmental Open Source Solutions
www.hydrologis.com

Andrea Antonello
Environmental Engineer
mobile:  +393288497722

"Let it be as much a great honour to take as to give learning,
if you want to be called wise."
Skuggsja' - The King's mirror - 1240 Reykjavik
____________________________________________________________________________