I learnt a techinque today for post-processing photos, and gives a real ‘pro studio’ look. It’s slightly over-emphasised in the following shot because I’m still feeling my feet with it, but for a taster, compare the tree and river shot from Saturday before and after:

Click on each for the full size version.

I can see how it might not be to some people’s taste, but subtly applied it gives a beautiful glow and softness without compromising realism. It works beautifully on people, too. I’ll play a bit more…

I just love the way cars come out in these model fakes. I wonder if it’s because we’re tuned to seeing model cars more than buildings and so on.

Anyway… this evening, I worked on a model of the bowling green in Queens Park, Loughborough, as taken from the Carillon tower on Saturday…

click for full size

1. Take a suitable source picture. Good things to look out for are:

• Strong lighting - models will normally be lit by a single room lamp. Strong shadows help the illusion.
• High vantage point - it’s rare that you take a photo from ‘inside’ a model, you’re normally looking down on it.
• Long lens - to compress the perspective.
• Large - make sure the resolution is large enough to work on. My source is 1000 pixels wide.
• Simple - the mind won’t accept it’s a model if there is too much going on, or too much detail.

2. Once you have your image loaded in photoshop, duplicate the layer and add a blur to the top layer. It doesn’t really matter how much right now, it’s only for visualisation. I used about 4 pixel Gaussian blur, I think.

3. Now we start getting complicated. Show the layer mask for the blurred layer and create a gradient blur giving black at the focal point and white at the extremes. This means that the blurred layer will be visible in foreground and background, and the sharp underlying layer will be visible only in a narrow plane.

The results can be seen below:

4. Now we start to really pop the model from its surroundings. Working on the layer mask, use a brush and the alt key to pick up the mask value at the ground level of any vertical objects, and paint over the object with that value. So… the base of the house is in focus - being on our focal plane, so we pick up the mask value (black) and paint it all over the front face of the house. Because the front face is in our focal plane, it should have full detail and not be affected by the gradient of focus.

Similarly, any trees or other tall objects should have the colour at their base picked up and painted over the whole object. The results start to look good:

And your layer mask should look something like this:

5. So far, we’ve only been working on our ‘preview’ blur. Now we’ll do it properly.

Duplicate the first (non blurred) layer and drag the layer mask onto it. Discard the ‘preview blur’ layer. We should now have two identical layers, but the one on top has our crafted layer mask applied to it (with no effect right now).

Now, apply a lens blur, taking the layer mask as the depth map source. That means a high-quality blur will be applied with defocussing matching the mask we have created. The settings might look something like the following, but fiddle around until you get a good effect:

6. Our results are now looking pretty good, but there’s a slight hitch. Any defocussed foreground objects will have sharp edges. See the bush in front of the left side of the house? It has blurred detail but sharp edges:

Simply use the blur tool on such objects until their outlines are pleasingly indistinct:

And we’re done.

Today, I mostly learnt how to turn a real photo into a model…