Skip to content

Improving on the Georeferencing

May 29, 2012

As a follow up to the previous post about pulling apart the old application, this post will describe the efforts that have been made by Edina to improve on the accuracy and quality of the Booth map image that will be utilised by the mobile application.

After a bit of digging LSE were able to supply us with scanned copies of each of the original 12 map sheets that make up the full Booth map. These new map sheets demonstrated a much higher resolution then the original map and were therefore clearer at higher zoom levels. Importantly for our purposes this gave us the benefits of improved readability of map based text (especially streets) and also an enhanced ability to discern between the colour shading of the streets that form the central part of the booth map.

The individual maps came as tiff files but in a more traditional presentation format and each image included a title, legend, scale indication (in furlongs) and a border around the map. For our purposes none of this information was required and as we wanted to join the different map sheets together, these additional items needed to be removed. Adobe Photoshop was utilised to crop all the images as close as possible to within the applied border and to smooth out the edges of the images.

Now that we had a copy of each of the maps in the correct style we needed to georeference them, both against an Ordnance Survey backdrop and also a Google Maps backdrop. The free and open source GIS package Quantum GIS (QGIS) was used for this as it has a plugin for loading a variety of web services into the map canvas – including Google Maps, OpenStreetMap, Yahoo and Bing. Additionally QGIS can host the Ordnance Survey web service Edina provides called OS Openstream, thus providing us with both our backdrops to georeference against.

More information on this can be found here:

QGIS Openlayers Plugin –

EDINA OS Openstream –

Georeferencing the full 12 sheets did take a while as each sheet required 15 or more control points and the process also had to be repeated for the two different map projections. The resulting map sheets fitted together like a jigsaw puzzle but because some of the maps covered parts of the same area there were large parts that overlapped. Again Photoshop was utilised to cut out certain areas of the map sheets to reduce the overlapping but also to fit the sheets together cohesively. The resulting new map instantly demonstrated a vast improvement on both the georeferencing and also the clarity at high zoom levels. We built a simple Openlayers demo site to compare the new map to the old map side-by-side and the results were pretty conclusive.

However, the new map did not come without problems. As a consequence of the improved georeferencing, unfortunately in some cases the sheet edges did not match particularly well. More effort could be spent on dealing with and mitigating this issue, but within the timeframe of the current project it has to be considered as a bit of a trade-off: a vastly improved map image in exchange for the slight miss match at some of the tile edges.


From → Uncategorized

  1. blair freebairn permalink

    Your map edge issue could have been avoided if you did the image cropping and stitching prior to the geo-referencing. If you look closely at the different sheets you will notice that where two sheets cover the same area the underlying mapping (stripping out the parish boundaries and lables) are identical. This means you could edge match the thematically shaded areas from one sheet to the other exactly simply by cropping, cutting and pasting. (Make sure you cut along pre-existing printed lines like a street edge and without transecting any text) I haven’t checked all 12 sheets but the couple I have checked at do seem to be on at the same scale. If any aren’t you could use the scale bar to zoom your image in or out till it is exactly matched.

    Only then once you have created your single massive mosaic image do you do the geo-referencing by assigning control points to exact locations based on satellite imagery. That way you get seamless edges and a single correct geo-referencing schema for your image.

  2. Whilst the above is a logical workflow orientated point of view it has to be remembered that we were a) working with largely unknown materials of uncertain provenance b) time constrained and c) looking to quickly explore how we could enhance the woeful state of the original single stitched map (which incidentally came as a large non-georeferenced TIFF). Warping is variable across the full extent of the map and highly localized in certain areas – it made sense to do a per sheet warping for expediency in road-testing – and then stitch. This appeared to be a less upfront resource intensive process – compare the current with the orginal and you’ll see that the current one is vastly improved. Ideally (and if your volunteering…), we’d redo on the linear workflow you suggest – or better yet we’d go back to the undigitised large scale hardcopies and digitise from scratch. Sadly, time and money mean our world is often less than ideal and I think what we got was a relatively quick, cheap(ish) and for our primary use case (delivery to smartphones), fit-for-purpose compromise.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: