Want to clean up water stains or other dirty metal surfaces in your cityscape and architectural images?  If you photograph modern buildings, you’ll inevitably run into this tough issue.  Cleaning up the metal can make the finished image look stunning, but can also be a painfully long process using standard cloning and healing techniques.  Thankfully, there’s a much easier way…

Here’s a quick summary of the process to clean metal surfaces in Photoshop:

1. Duplicate the image and apply the Surface Blur.  This will blur the metal surface, and keep natural looking edges (if the metal blends with the surrounding areas, we’ll get funny effects at the edges).  The goal here is to remove as much dirt as possible, but you probably can’t get it all without starting to create other problems.  That’s ok, just do as much as you can in this step and we’ll get the rest in the next step.
2. Create a new blank layer above the surface blur layer.  Use the clone stamp to fix remaining dirty areas that couldn’t be removed with Surface Blur.  Use a lower opacity to help blend as you work.
3. Add a 50% gray layer above the clone layer, set it to overlay blend mode, and then Filter/Noise/Add Noise.   This will add back “grain” to the image that was lost in the previous steps.  This will help the retouched areas blend with the rest of the image.  Reduce the opacity of this layer as needed to get a natural result.
4. Put these three layers into a group, add a black layer mask, and paint white over the areas where you want to clean metal.
5. If needed, put the group into another group with a white mask and black out areas you want to protect.  Using a combination of masks can make it much easier to use selection tools (such as luminosity masks or the magic wand) to protect certain parts of the image.  And using a non-destructive workflow gives you more flexibility to fix any mistakes you find later.

In the video tutorial above, I created three group masks to control which parts of the image were cleaned up.  Using Lumenzia‘s “Combine” function, I’ve merged the masks from the tutorial into the single mask shown below.  This new mask is the functional equivalent of the three masks I created (but it’s a lot easier to understand what’s happening when you view it as one mask instead of three).  As you can see, there are a few areas that look sloppy.  Some of the sky is included in the mask (since I didn’t both to black it out).  The bottom right has a messy hand-painted area to paint in some missing parts in the mask.  And there are a few other funny looking areas.  None of these matter, because these are areas where the surface blur has no real detectable impact.  And that’s one of the benefits of this approach.  You don’t have to be overly precise, so you can work more quickly.  The surface blur and masking technique helped protect the image where needed, but we don’t often need to create a perfect mask to get an excellent result – so there’s no need to waste the time more precisely refining this mask.  If this image was to be printed (instead of just being used as a quick demo), I would put a little more time into a few of the edges – but it’s already very close to the result I would need.