Horizontal and Vertical Constraints
Once your constraint is placed in its final position, right-click the cyan line (it will be green if you rotated it first) and choose Vertical or Horizontal, depending on your situation (Figure 7). When you do this, Photoshop CS6 will transform the area immediately surrounding your constraint in order to “flatten” the perspective and create an edge that is horizontal or vertical, relative to the edges of the frame.
Figure 7: Setting a constraint to a value of “Vertical” or “Horizontal” by right-clicking on it will tend to flatten the perspective and “square up” the pixels in that area.
Once you have familiarized yourself with the process of creating Constraints, the rest is easy: find additional lines in your image that need constraining, and follow the same process noted earlier. Typically for any type of complex architectural photo, I find that I may have a dozen or more constraints in total before I’m done; how many is really dependent upon the complexity if the composition. For this example I added four vertical Constraints to handle the central portion of the image (Figure 8).
Figure 8: Adding additional Vertical or Horizontal Constraints across your image can go a long way toward achieving your final look.
The next step for this example was to determine if there were any good horizontal references. There were only two or three of them, and they were much shorter than the vertical ones; the curvature of the architecture prevented me from using the underside of the walkways as a true straight-line reference.
Instead I used some of the first floor doorframes (background) and window framings, as well as two banks of lights on the 2nd and 3rd floor (Figure 9). These were not precisely perpendicular to the angle of the lens but they were close enough that using Horizontal Constraints on them produced a good result (yellow lines). You may need to experiment in situations like this to get the best result. Different numbers and types of constraints in proximity to one another can produce very different results.
Figure 9: This screenshot shows both Vertical and Horizontal Constraints acting on the image, straightening its curved lines and flattening the perspective.
If you find any grid-like patterns in your image, applying constraints to those can help to further smooth out your photograph. In this case, many of the ceiling tiles were curved from the camera’s perspective. I flattened those areas by adding new Unfixed Constraints and following the lines between the tiles, which I know to be straight lines (Step 10).
Step 10: Using Constraints to straighten out polygonal or rectangular objects can help to further smooth or “flatten” the perspective in the image.
Once you feel that the image has been straightened out sufficiently, you can get a better view of things by turning off the Constraint widgets. At the bottom of the window, deselect Show Constraints, then zoom out so the whole image is visible (Figure 11).
Figure 11: Previewing the final corrections with the Constraint widgets turned off can make it easier to evaluate the image before processing the file.
If you like the result, click OK and the GPU-accelerated Adaptive Wide Angle filter will process your image in just a few seconds, sending your final output back to the canvas for cropping or content-aware fill edits (to fill gaps left behind). In this case because edits were made across the image, quite a bit would have to be cropped away along the edges. However if you account for AWA when shooting the picture, leaving some extra space around the edges of the frame, you’ll be in good shape as most sessions would not require as many transforms as this one did.