Jump to content

 
Photo

Multidirectional Hillshading Tries

- - - - - hillshading mdenoise ASTER

  • Please log in to reply
2 replies to this topic

#1
Frank B.

Frank B.

    Newbie

  • Validated Member
  • Pip
  • 7 posts
  • Luxembourg

Hi,

 

while I am waiting for my DEM creation to finish (see this post)  I am playing a bit around with different hillshading methods in order to obtain a better understanding and some visual guides for my self. This post doesn't contain directly a question but some of the results and insights I have gained so far. I hope they are also helpful to others. Of course I am always interesting in the feedback from other cartographers, so feel free to comment!

 

For another project of mine I am currently doing some terrain analysis in Cape Verde, more exactly the island Santo Antao. The island has a substantial height above sea level (1979 m) and the surface is formed by volcanos and shows deep, sharp valleys and landforms. The surface is really quite complex so I think Santo Antao is a great candidate for some hillshading tests.

 

I don't have a good source for elevation data so I settled with Aster satellite DEM. Aster data has a nice horizontal resolution of 30m per pixel, but tends to have errors and anomalies in the form of pits, little hills and so called "mole runs". Using this data as it comes is rarely adviceable and postprocessing is necessary, also in order to gain some form of generalization for the hillshading.

 

There is a nice little tool called mdenoise which seems perfect to remove the noise in satellite DEM in general. In combination with GDAL it is pretty trivial to apply the denoising algorithm on the ASTER DEM. There are two parameters which control the amount of smoothing applied to the scene: The number of iterations over the DEM and a treshhold which preserves sharp features. I've written a small script [see below] which created denoised scenes ranging from 5 to 30 iterations and threshold values from 0.7 to 0.99. I used GDAL to create a standard analytical hillshading and a hillshade/slope hillshade from the denoised DEM in order to gain more insight on the result. All this together (1 DEM and 2 hillshades per denoised DEM) created 2341 files with 8.3 GB :D

Fortunately I didn't have to regard all of these - it became quickly evident that 15 iterations and a threshold of 0.7 held a good balance of smoothing and preserving terrain features. If there's some interest in this I am willing to create jpegs from the hillshades and upload them onto my server so that everybody can choose his own favorite.

 

Anyway, I've attached several images to this post. The black and white ones are the hillshade tries. The first is a standard hillshade with 315° and a high azimuth like Google, I chose 70° as a value. The next two are more complex in their generation. In order to pay tribute to the relative complex nature of the surface I used a MDOW (Multidirectional oblique-weighted) hillshade approach. I calculated analytical hillshades for 225, 270, 315 and 360 degrees and used GIMP to combine them together. I loaded them as layers into GIMP and gave each a layer transparency of 25% and used the multiply blending mode. In another try I added a slope map to the MDOW hillshading but the differences are very subtle compared to the hillshade without slope. Please refer to the files names to see the creation options for the hillshade.

 

In order to achieve another measurement to see how well the hillshade performs I downloaded a Landsat 8 scene (which has btw. the same horiz. resolution like the ASTER data) and pansharpened the Landsat scene. I overlaid the hillshade on top of the Landsat scene in Quantum GIS and chose again multiply to blend the hillshade on top. In order to see how well the MDOW hillshade compares to the analytical 315° hillshade I made combinations of all of these and you can find the corresponding results attached below.

 

I like how subtle the the MDOW hillshading is in contrast to the analytical hillshade. I think it helps in the right places and is yet invisible where it isn't necessary. Right now I am doing a hypsometric tint of the DEM but haven't found a good color scheme yet. I want one that follows the vegetation (brown at the bottom and green in the middle to top heights) but I haven't found a good approach yet. As soon as this is done I will share the result here too.

 

Well I hope you found the little excursus interesting,

 

Frank

 

Here's the script

#!/bin/bash

IN="santo-antao_lanczos.asc"
DEM="santo-antao"
EXT=".tif"

MDENOISE="/home/frank/bin/mdenoise"
TRANSLATE="/usr/bin/gdal_translate"
GDEM="/usr/bin/gdaldem hillshade -z 3 -alt 45"


for n in $(seq 5 30)
do
    for t in $(seq 70 99)
    do
        BASE="${DEM}_n${n}_t$t"
        DENOISED="${BASE}_denoised"
        $MDENOISE -i $IN -n $n -t 0.$t -o $DENOISED.asc
        $TRANSLATE $DENOISED.asc $DENOISED$EXT
        $GDEM $DENOISED$EXT ${DENOISED}_hillshade$EXT
        $GDEM -combined $DENOISED$EXT ${DENOISED}_hillshade_combined$EXT
    done
done

Attached Files



#2
david17tym

david17tym

    Key Contributor

  • Validated Member
  • PipPipPip
  • 99 posts
  • Gender:Male
  • Location:North Wales
  • United Kingdom

Hi Frank,

I've been using the MDOW approach too for my hillshades, using this ArcGIS resource:

 

http://mappingcenter...s.modelsScripts

 

"Hillshade Tools" - overdue an update to v10 if anyone from ESRI is listenning.

 

Lovely images of Santo Antao.

Dave



#3
Matthew Hampton

Matthew Hampton

    Hall of Fame

  • Moderator
  • PipPipPipPipPipPipPip
  • 1,295 posts
  • Gender:Male
  • Location:Portland, Oregon
  • Interests:Playing in the mountains and rivers.
  • United States

Very nice work Frank.

 

I like the MDOW greyscale hillshade the best, but almost prefer the straight 315 when multiplied with the imagery.  The MDOW get's overwhelmed by the intensity of the imagery.  If you made the imagery a little transparent then multiplied the layers, I think the nicer hillshading would show through better.

 

I'm intrigued to see how the hypso-tinting works out for you.


co-cartographic creator of boringmaps.com






Also tagged with one or more of these keywords: hillshading, mdenoise, ASTER

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

-->