Remote Sensing Tools

Typically, pan-sharpening in Remote Sensing / GIS has been considered the domain of those programs and applications. But, the reality is — that pan-sharpening is a fairly simple process to achieve in a program such as Adobe Photoshop®, if you have the appropriate processing algorithms.

The following is a typical procedure for pan-sharpening using the Convolution algorithms provided with Remote Sensing Tools.

Preparing the High-resolution Panchromatic Source Image

First, load your high-resolution panchromatic image and convert it to RGB.

1. Click through Image → Mode → RGB Color, to convert the pan image to RGB color from Grayscale.
2. Set the image aside in the work-space.

Preparing the Low-resolution Multi-spectral Source Image

Now, load your Multi-spectral color image and resize it to the size of the Pan image. In this example, I’m using the resize value based on an image that’s 0.6 or 1 meter resolution, respectively, such as imagery from Digital Globe, or GeoEye. If using Landsat ETM+ 14.25 meter imagery, then 200% would be the appropriate resize value.

1. Click through Image → Image Size.
2. With Constrain Proportions checked, set the Document Size to 400% in either width or height. Ensure that the pixel proportions shown in the upper fields are consistent with the pan-image’s dimensions.
3. Click OK to resize the image.

Assemble the Image

Copy the Multi-spectral image and paste it into the Pan-chromatic image.

1. Click through Select → All, (or Ctrl + A) to select the entire image.
2. Click through Edit → Copy, (or Ctrl + C) to copy the Multi-spectral image to the clipboard.
3. Now, click the Pan image’s document window to enable it again.
4. Click through Edit → Paste, (or Ctrl + V) to paste the multi-spectral image into the pan-image document as a new layer (Layer 1).

Reorder the Layers and Process

1. Double-click on the Background layer (the pan-image layer) and click OK at the dialogue, to release it to a transparency-enabled, movable layer. (Or, right-click and choose Duplicate Layer to generate a copy of the Background layer that can be moved.)
2. Move the pan-image layer to the top of the layer order.
3. Click through Remote Sensing Tools - Convolutions → Centre Weight Highpass - 8 to run that process. (Depending on the size of the image, this may take a couple moments to process.)
4. Once the processing is complete, set the layer to Linear Light.

That’s it! You should now see a pan-sharpened image that retains all the characteristics of the multi-spectral image, without compromise or major loss of radiometric value of the multi-spectral source!

As a final step, you can decide to run the Flatten Image command at this time, or you can chose the additional alternative steps which I’ll throw-in as an enhancement possibility out of my own experience in working with this type of imagery.

Optional Steps to Enhance Image Quality

One of the issues that comes from pan-sharpened imagery is that the multi-spectral image retains the common effect of ‘ghosting’ or ‘color-haloing’ — which is a direct result typically found in moving objects such as vehicles. This is simply the lighter-blotchy areas that might occur in some or most imagery where pan-sharpening is used as the fusion technique.

In Photoshop, there is a way to dramatically reduce this effect, though it does compromise some of the multi-spectral features — such as asphalt (roads) and large bodies of water, for example. Overall, however, it’s a decent methodology to help enhance a pan-sharpened image using the pan-chromatic source.

Let’s back-track then through the steps, and if you’ve generated an additional duplicate of the Background layer (the pan-image layer), then you can do the following.

1. Move the additional duplicate Background layer (duplicate of the pan-image) to the top, above all other layers.
2. Set this copy layer to Pin Light.
3. Now, right-click the multi-spectral image layer and create a Duplicate Layer of that specific layer.
4. Move the duplicate multi-spectral layer to the top, above all other layers.
5. Set this new layer as Color.

That’s it! What you may have noticed, was that the duplicate Pin Light layer of the pan-image had drastically reduced the ‘ghosting effect’ around the tiny cars and other features, as well as take care of some of the ‘blotchiness’ that occurs in roadways or shadows. But you may have also noticed that it appears that grey values may sometimes replace very small areas. This is why we’ve made an additional duplicate copy of the multi-spectral image, in order to re-color the entire image, to ensure that we’re getting the best of both worlds.

Again, the optional method is what I consider an enhancement alternative toward a final image. But keep in mind the caveats associated with altering the original radiometry of the multi-spectral source. In most cases this may be a desired step to take, but in some cases it may not. It really depends on what your end-result has to be and toward what application you’re producing imagery for.

In most cases, it’s been a technique that I developed that’s helped in processing some rather stunning imagery for representational or fusion-type applications — and it’s also been accepted in even the most strict of application environments, such as vis-sim, where material-classification is more dominantly used to alter materials within the visualization environment.