Archived posting to the Leica Users Group, 2008/03/11
[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]Here is a way to remove "tourists", cars, etc., using Photoshop CS3 Extended, without using the clone tool. It works really well in situations where the things you want to remove are moving, not so well otherwise. When it works, it seems like magic. You can make a freeway appear empty. 1. Take a series of shots. Raw or jpg doesn't matter (but there will be an extra step to open the raw images) Tripod? A tripod is useful here, but if you can hold the camera reasonably steady, you will fix the alignment in step __. If you can eyeball some reference (like a window that will be centered) you can even put the camera down between shots. Number of Shots? This will depend upon the movement of the tourists. Try 10 to experiment with the method. For mathematical reasons (which may or may not become clear to you, but don't worry), more than half of the shots must show background for each pixel to guarantee removing the tourist. So, if you have one tourist who won't move, you won't be able to remove him no matter how many shots you take. For cars on a freeway, for example, 5 to 10 shots are needed. More is better. Frequency of Shots? This depends on the speed of the things to be removed. Ideally, each shot will have a tourist in a different position, so that background is being revealed at least half the time. Start with around 1 second intervals to experiment. 2. In CS3 Extended, choose File>Scripts>Load Files into Stack ... Browse to the location of your shots, and be sure to check the two boxes for "Attempt to Automatically Align Source Images" and "Create Smart Object after Loading Layers." Which Shots to Select? Suppose that you took 15 shots. Unless you have a speedy computer, choosing all 15 shots means that you can now go have dinner. I recommend choosing only three to five shots to start, being the first, last, and some middle shots (use Command-click). 3. When the program finishes processing, choose Layer>Smart Objects>Stack Mode>Median. Most of the tourists should now be gone. If not, repeat with more shots in the stack. As I said, when it works it seems like magic. How does this work? For each pixel, a stack is created with the luminance values from each shot. Suppose that we have 15 shots, and that value 128 is background, value 25 is a dark shirt, value 200 is a white shirt. The first few shots a fellow with a white shirt is in the photo, then he moves and we get some background, and then someone with a dark shirt is there for a few shots. The stack of 15 shots might show these values: 1 200 2 200 3 200 4 128 5 128 6 128 7 128 8 128 9 128 10 128 11 25 12 25 13 25 14 25 15 128 Now, the computer ranks the 15 values in numerical order: 25 25 25 25 128 128 128 [128] 128 128 128 128 200 200 200 The median in this list is the middle value, in this case the one I put in brackets [ ]. That becomes the output value for that particular pixel, and as you see it is conveniently a background value of 128. Now it should be apparent why to guarantee removing the tourist, more than half of the values need to be background. Regards, Bob Rose