Clearer mountain views


I have been working during the past few months on a number of technical improvements for better quality in 3d views. These i want to introduce here together with a number of images from the Kashmir/Karakoram/Pamir region which is well suitable to demonstrate these enhancements.

Removing shading

All satellite images as taken feature the specific illumination due to the position of the sun at the time the image is taken. If you visualize the satellite image directly the shading due to the directed lighting is essential for reading the image. The flexibility in choosing a specific illumination in most cases is limited to selecting between different times of the year although in certain situations you have more freedom as i have written on several occasions here in the past.

Most people who produce 3d visualizations based on satellite images use this inherent illumination as recorded as the basis of their visualizations. This leads to a complete lack of flexibility with regards to illumination and also to subtle inconsistencies because the shading visible does usually not accurately match the 3d geometry shown and even subtle differences here lead to a lack of realism being perceived by the viewer and distracts from the actual content of the visualization.

Therefore my 3d views feature a precisely calculated simulation of the illumination individually selected for the view that is independent of the illumination when the image data used was recorded. Many of my views for example show an evening lighting although the satellite images used are recorded in the morning. For this to work well i of course have to remove the shading effect from the image data. I have been doing that for more than ten years now. The process for this has been refined significantly over the years and the most recent improvements in particular led to better accuracy and more robust dealing with difficult situations. Here an example.


Original L1C image

Atmosphere and shading compensated

As you can see the shading compensated version looks annoyingly flat and structure-less – but this is exactly what it is meant to look since the actual impression of the earth surface topography is to come from the specific simulated shading calculated on top of this in the 3d rendering. And while you can’t see the relief structure any more looking at the image the actual differences in surface coloring are better visible after the shading is removed.

Clearer geometry

The other improvement i have introduced here is a new processing of the geometry data to reduce noise while preserving acuity. I made use of the ALOS AW3D30 relief data – which offers fairly good coverage in the region in question. This data – like all other similar data sets – features a significant level of uncorrelated noise which is well visible when you use it in rendering directly. Reducing this noise while preserving the actual relief features leads to a clearer and better readable rendering. The technique used is related to the methods i use for producing generalized shaded relief rendering in 2d.

Original elevation data with noise

With noise reduced and generalized relief data


Here are various examples from the larger Kashmir/Karakoram/Pamir region rendered using the techniques described. All of these and more can be found in the catalog.

First for comparison the iconic K2 view i show on the main page on in its old and new version.

K2 view from 2006

K2 view from 2019

And here a selection of further images of this region.


  1. What algorithms do you use for topographic correction of satellite images and for DEM noise filtering? I would like to know technical details.

    • There are no details on these methods publicly available at this time. They were developed for in-house use here, data processing can be performed as needed for customers.

Leave a Reply

Required fields are marked *.

By submitting your comment you agree to the privacy policy and agree to the information you provide (except for the email address) to be published on this blog.