[gdal-dev] gdaldem definitions

[Even Rouault]

Le lundi 22 septembre 2014 01:45:16, Dominik Schneider a écrit :
> One correction, that should have said:
> TPI "the central pixel is assigned the absolute difference between the
> central pixel of a 3x3 window and mean of all pixels within the 3x3 window"

Domink,

here's the relevant part of the code. Looks they match the definition although 
I guess there's some ambiguity sometimes if the central pixel must be take 
into account or not. The main difference between the implementation and Wilson 
article might be on TPI where GDAL takes the difference and not the absolute 
difference. AFAIR, the implementation in GDAL is similar to the one in GRASS, 
but there might have been changes during porting.

Pixels are numbered like the following :
[0] [1] [2]
[3] [4] [5]
[6] [7] [8]

so [4] is the central pixel.


/************************************************************************/
/*                         GDALTRIAlg()                                 */
/************************************************************************/

float GDALTRIAlg (float* afWin, float fDstNoDataValue, void* pData)
{
    // Terrain Ruggedness is average difference in height
    return (fabs(afWin[0]-afWin[4]) +
            fabs(afWin[1]-afWin[4]) +
            fabs(afWin[2]-afWin[4]) +
            fabs(afWin[3]-afWin[4]) +
            fabs(afWin[5]-afWin[4]) +
            fabs(afWin[6]-afWin[4]) +
            fabs(afWin[7]-afWin[4]) +
            fabs(afWin[8]-afWin[4]))/8;
}


/************************************************************************/
/*                         GDALTPIAlg()                                 */
/************************************************************************/

float GDALTPIAlg (float* afWin, float fDstNoDataValue, void* pData)
{
    // Terrain Position is the difference between
    // The central cell and the mean of the surrounding cells
    return afWin[4] - 
            ((afWin[0]+
              afWin[1]+
              afWin[2]+
              afWin[3]+
              afWin[5]+
              afWin[6]+
              afWin[7]+
              afWin[8])/8);
}

/************************************************************************/
/*                     GDALRoughnessAlg()                               */
/************************************************************************/

float GDALRoughnessAlg (float* afWin, float fDstNoDataValue, void* pData)
{
    // Roughness is the largest difference
    //  between any two cells

    float pafRoughnessMin = afWin[0];
    float pafRoughnessMax = afWin[0];

    for ( int k = 1; k < 9; k++)
    {
        if (afWin[k] > pafRoughnessMax)
        {
            pafRoughnessMax=afWin[k];
        }
        if (afWin[k] < pafRoughnessMin)
        {
            pafRoughnessMin=afWin[k];
        }
    }
    return pafRoughnessMax - pafRoughnessMin;
}

Even

> 
> Dominik Schneider
> o 303.735.6296 | c 518.956.3978
> 
> 
> On Sun, Sep 21, 2014 at 5:36 PM, Dominik Schneider <
> 
> dominik.schneider at colorado.edu> wrote:
> > Hi -
> > I was reviewing the gdaldem documentation and the definitions provided
> > seem to differ from those in the cited article (Wilson et al. 2007)
> > For example,
> > TRI should be "the central pixel is assigned the mean of the absolute
> > differences between a central pixel and its surrounding cells"
> > TPI "the central pixel is assigned the absolute difference between the
> > central pixel of a 3x3 window and all pixels within the 3x3 window"
> > roughness "the central pixel is assigned the largest inter-cell
> > difference within the 3x3 window surrounding the central pixel"
> > 
> > of course, this is assuming the algorithms are actually those presented
> > in the publication at the 3x3 scale as opposed to those written in the
> > documentation.
> > 
> > I assume the source code could be checked to resolve this question?
> > 
> > Thanks and my apologies if I misunderstood something.
> > Dominik

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com