after generalizing my DHM and extraction of different color-hillshades, I got something like this:


Its not very good, but I used the idea of Imhof generalizing and coloring the hillshade (multishade). Its "digital-round-smooth" like "organic" and not "natural-edged-hard" like I want it to. (Sorry for that kind of english :)

I mean, in natural its a very steep and hard mountain range, and not a hilly "hobbit-land" :)

Do you have any Idea, how to construct it more natural?

Thanks! Martin

  • 2
    Are you sure you didn't overdo the smoothing in gis.stackexchange.com/questions/24348/…? Can you describe your workflow so far so we can reproduce it? – underdark May 4 '12 at 10:15
  • 1
    Yes, this DEM has been grossly over-generalized, rounding off all ridges (and the bottoms of the valleys). No cartographic trick is going to recover the sharpness you seek. You can keep the generalized DEM if you like and still recover the appearance of sharp ridges by hillshading the original DEM and overlaying that hillshade with a semi-transparent rendering of the generalized surface. – whuber May 4 '12 at 14:38
  • @underdark: In this example I did, yes, just do show it in a extreme way. I have other examples, but there also still other problems... I tryed, and tryed and tryed... and finaly I think, the best is to use r.denoise but i cant get it running (gis.stackexchange.com/questions/24854/how-to-install-r-denoise) – MAP May 6 '12 at 12:30
  • @Map, I realize you asked this long ago now but thought I'd add that the most unusual thing about this map is that it appears that there are many illumination sources rather than the tradition NE for hillshades. Notice that there is shadow on N, S, and even E and W facing slopes. – WhiteboxDev Sep 11 '14 at 22:24

I usually post-process a copy of the raw hillshade in some external graphics editing program that I can directly over-write (save, instead of save-as)

Photoshop and GIMP are two that come too mind.

The process in a nutshell I follow: Import GeoTiff'd Hillshade -> Tone -> Median -> Reduce Noise -> Gaussian Blur

  • Your geotiff should be accompanied with a worldfile (like .tfw) for portability between the GIS and graphics program you use)

Use the tone tool to redistribute your min/max values of the greyscale hillshade. In GIMP you use the 'Levels' tool. I often use a very small range with a min of 5% grey and 25% grey. Rarely do I use full white to black greyscale ranges.

Median is used to generalize. This is the part where can easily misrepresent your hillshade. In GIMP the equivalent tool is 'Despeckle'.

The median/despeckle tool will reduce the sharp edges and make then rounded, like mountian tops and valleys. Thus often making it appear as though some valleys are filled-in and some mountain tops are flatter than they appear. Use this tool lightly, or not at all.

Reduce noise is used to remove any strange artifacts that median couldn't - often I don't even use reduce noise. Almost like a cheat way of de-striping your DEM if it suffers from 'stepping' - a result of manual profiling.

Gaussian blur will introduce the final smoothing of the hillshade and is often the last thing I do (this will also help remove any introduced artifacts).

The other option, is to run the above process over an HDR hillshade.

Create hillshades from various aspects like 360 (or 0 deg), 15, 30, 45. Those are the four angles I might use for my local region because that is the general direction the sun comes from during the morning to mid-day. HDR the four images using either GIMP or photoshop then use whatever generalizing filters you want like the aforementioned ones.

Experiment and see what works best for you.

What's good about working on a GeoTiff is that none of the georeferencing is lost as the pixel size and extent are not altered in the process.

There are lots of other filters/overlays to use and play with. The ones I've mentioned are the ones I use most frequently. Also remember that the less filters you have to apply the better so that the terrain doesn't become dramatically misrepresented.

When I apply the filters, they are done so in small doses. For instance, I find myself never applying a gaussian blur over 5%.

Results from 50k GeoBase.ca DEMs Useful resources:


Cartographic (not GIS) hillshade techniques used not to be straightforward. The swiss cartographic school applies a computer assisted method but good results can be obtained using the resolution bumping method (sometimes works very well and gives you a really better model than using traditional hillshade algorithm). Please take a look at: http://www.shadedrelief.com/dem/dem.html


Based on the image that you included above, and the image you had of your original data from a different post, I think you are encountering the pitfalls of generalizing the raster.

First, I will reference a link that was given in comments on one of your previous posts. It is a discussion by @underdark about Generating Relief Maps in QGIS. She is talking about using a Raster Terrain Analysis plugin for QGIS. What is important is that she shows how the elevation bands are set to a specific color, and also how the Z-factor affects the smoothness of the relief.

Both the changing of the color bands and changing the Z-factor will generalize your data.

I think the Z-factor can have a much more dramatic effect and it is more important that you get this set correctly based on the units that your x,y coordinates are in compared to your z coordinates. Essentially, if your horizontal and vertical units are the same, then your Z-Factor should be 1. Increasing it in this situation will cause potentially dramatic smoothing of your data, similar to what your photo shows above.
The Number of color bands you choose is the next part of the generalization process. The fewer bands you have, the more the amount of relief will be reduced.


What follows should be performed on a copy of the original data, not on the generalized version. Any sort of smoothing or generalization process degrades the data, so if there is a problem, you don't fix it , you start from scratch. This also makes the case for working on a copy of the original, so you have a backup.

So, in order to increase the texture of your DHM, you need to first make sure your Z-factor is set properly if you are using the plugin mentioned above.

Next, I would increase the number of color bands in the hillshade. More color bands means smaller ranges, which leaves more texture in the hillshade. This will retain more of the ruggedness and sharp edges in your hillshade.

Hope this helps! Russell

  • 2
    I believe that changing the Z-factor will not appreciably change the strongly rounded appearance of the ridges without causing unwanted side effects (high contrast and cliff-like appearance of slopes). Have you tried out your solution? Can you post examples of it? – whuber May 4 '12 at 14:41
  • 1
    @whuber, I will edit my post to reflect this comment, but to be clear, I think any changes to z factor should be done based on the original dataset. Doing any further operations on his derived dataset will only cause further distortion, as you said. – Get Spatial May 4 '12 at 18:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.