or Connect
New Posts  All Forums:Forum Nav:

Photo Montages

post #1 of 31
Thread Starter 
Does anyone know how to generate photo montages? I'm talking about the kind where there is a fixed background of a race course with multiple images of the same racer making a turn around a gate.

Do you take multiple exposures on one frame or multiple frames?
What time interval is used between exposures?

How do you keep the skier properly exposed without overexposing the background if you do one frame with multiple exposures or do you fix the background by post processing?

Is specialized software needed (other than consumer level photoshop)?

Is there a description published anywhere for the entire process from picture taking to image processing?

Thanks in advance for any advice.
post #2 of 31
Are you asking for something like this?

My suggestion is to shoot multiple frames of the same subject and then combine them using multiple layers in Photoshop. The key concept is the transparency mask. Your bottom layer provides the background and all other layers have only the skier visible (i.e. everything else is made transparent). it is a bit of a tedious job to paint the masks, so having featureless snow backgrounds like in this image helps, since you can use MagicWand tool to speed up the proces. A few oher tips: Shoot on manual so your exposure will be consistent from shot to shot; don't zoom in too close, otherwise you won't be able to use one frame for background (it is OK to stitch several, but will mean more PS work).

Do post your results here....
post #3 of 31
I did it algorithmically in Matlab. It was mostly an experiment. It came out pretty good, and I think I could do it a lot faster on a second project. Obviously if you are not a math or programming geek, this won't work for you.

Having done it semi-manually, I thing a software package could be written to automate the process. I suspect one might already exist.
post #4 of 31
Having done quite a few of them, I can tell you that it is time-consuming and tedious, but I learn more about it with each montage I put together.

It starts with the original photography or videography. Most of mine come from digital video, although a few have come from my DSLR. A high-definition video camera, especially one that shoots in "1080p" progressive scan mode, will produce very good still images, depending on how it's set. Next best is a high-end standard definition camcorder that can shoot in progressive scan mode("frame mode" is what my Panasonic GS400 calls it). If you use a standard consumer-grade digital video camcorder, you'll need to "deinterlace" each frame to get even "OK" results. (You can do this with Photoshop, and with most better video editing software.)

Still cameras can do a very good job, of course, but they are limited in frame rate. 3 frames per second is sufficient only for low speed skiing. 5 or 6 fps is sufficient for most things. But video cameras shoot 15 (Sony progressive scan SD camcorders like the venerable TRV900), 25, or 30 frames per second. Not only is that plenty fast, but even if you only need 3 to 5 fps, you can choose exactly which moment to start from.

You want the cleanest, sharpest images you can find, so you want to shoot at a high shutter speed to freeze the action--even for video. Pro camcorders let you adjust the shutter speed manually--I would use nothing slower than 1/500 of a second. Lower end camcorders usually have some sort of a "sports mode" that freezes action pretty well, at least in bright sunlight.

You want every frame to be exposed the same, so if you are good at setting exposure, set your camera or camcorder to manual (if it's an option).

Shooting, composing, and zooming video for still montages is different from shooting for good video. In addition to setting the high shutter speed and shooting in progressive scan mode (both of which make the video a little jerkier), you want to make sure that no parts of the skier leave the frame at any point. And you must not zoom the lens during the sequence you intend to create the montage from. But you do want the image of the skier to fill the frame as much as possible. It's a hard balance to reach sometimes! It's OK if the skier moves around a little in the frame (lousy for video, but largely irrelevant once you mask the image from the frame). But you must not tilt the camera during the sequence. None of these "errors" are impossible to overcome with some careful Photoshoppery, but it will take much longer, and the results won't be ideal.

Once you have the images or footage you want, you must "capture" the frames to your hard drive. If it's video, you'll need some sort of editing software. I use Adobe Premiere Pro these days, but you can do the job with less sophisticated software. It will be painful, though, if you have to capture each frame one-at-a-time by hand. Premiere will convert a clip into individual frames automatically, using any frame rate you want. It will also deinterlace the frames if needed. Be sure to switch off any "frame blending" or other motion-blur-creating functions.

So now you have your frames, saved as a folder full of .tif, .gif, or other image format files, or as a single "filmstrip" image (my preference) from which you can copy and paste individual images, and you're ready to create the montage. The theory is simple--using Photoshop or equivalent, you'll start with the most-distant frame, and then cut and paste the closer frames on top, in sequence, one by one. Start with a blank canvas big enough for the entire montage--although you can enlarge the canvas whenever you need to. Copy and paste in the first image, complete with background.

There are several ways to do the subsequent frames. You can cut the skier image from the background and then paste just the cutout image into the montage. (If you do this, be sure to leave some fixed point from the background intact--a gate, tree, chunk of snow, or something else that appears in both images--so you can register the second image on exactly the right spot.)

Alternatively, you can paste the whole second image on top of the first, then cut away just the parts you don't need. To do this, set the top image transparency to around 50%, so you can see the image below it. Line up the backgrounds (something that doesn't move). Then, with masking tools or the eraser tool or both, cut out the background from the top image where needed to let the lower image show through. Leaving the background everywhere else allows the background of the montage to "grow" as needed. Remember to reset the transparency to zero (100% opacity) when you're done.

(Save your work often!)

Repeat for each frame, until done.

(Save your work often!)

Depending on the number of frames and the resolution of the images, it can become a very large file. If your computer starts bogging down, you may need to "flatten" the image as you go along, combining the individual layers into one. Be careful, though--once done and saved, you can't go back and make adjustments.

(Save your work often!)


That should get you started. You'll figure out and stumble upon all kinds of little tricks that will help improve the quality and reduce the time somewhat. But any way you look at it, it takes a while! It can be tedious, meticulous work, and there are no real shortcuts that I know of.

There is, of course, specialized software available that can automate the process to a degree. "Dartfish" is the program that creates the instant montages you've seen on some television broadcasts, particularly of the Olympics. It makes me sick how quickly and effortlessly it does its thing, but the results I've seen, while looking fine on television, are not of the quality of a hand-made montage. Still, I can't help but envy it! It is not inexpensive.

Good luck! Please post your results, as well as any good "tricks" you come up with.

Here are a few of mine, all of which have been posted here before:

Patrick Deneen, US Ski Team (moguls), training at Mt. Hood last summer; from 35 mm DSLR

Freestyler Sammy Carlson, also at Mt. Hood, from high-definition video

Cyprien Richard (France), Noram GS at Keystone last season, from high-definition video

Me, Arapahoe Basin, 2007, from high-def video

David Oliver at Telluride, from progressive scan standard definition digital video

Dan Egan, Arapahoe Basin 2008, from hi-def video

Ben Atkinson, Keystone A51 Terrain Park, from progressive scan std def video

Me again, Arapahoe Basin, from high-definition video

Toni Sears at Arapahoe Basin, 2008. Note that the camera was zoomed for the last frame, requiring that it be enlarged to match the rest of the sequence, and losing some sharpness. Don't zoom when shooting for montages!

Some of these images have some additional Photoshop work, to fill in the background (mostly through "cloning") and otherwise polish them a bit.

Best regards,

PS--Please note that these are all copyrighted images, and I reserve all rights.
post #5 of 31
I use photoshop, but not very well. These take about 30 minutes to produce each montage...

post #6 of 31
"But not well"? They look good to me, Fox!

Were those done differently from mine? Did you use a fixed camera on a tripod or something? That's another way to do it, especially if you're using a high-resolution still camera that produces clean images even when the skier takes up only a small portion of the viewfinder. It has the real advantage that you don't need to worry about registering each frame differently, or stitching together the background.

Best regards,
post #7 of 31
Bob, they were hand held using my Canon G9, which is about 2fps in continuous mode. The good thing is that I'm not taking pics of racers, so I can get away with that.
The images are at full size for the camera (4000x3000), and by keeping fairly still, there's maybe only 2 backgrounds to paste together.
post #8 of 31
Bob, you got some great results from video. Pretty cool stuff.
post #9 of 31
Thread Starter 
Thanks everyone. This is exactly what I was looking for. You guys have some great results. I have a copy of Matlab and a copy of the image processing toolkit. I use Matlab extensively at work but I'm embarassed to say I didn't think of using it for this application. I'll have to think about how to do this.
post #10 of 31
Thread Starter 
I think I have a way to automate the creation of the montages with Matlab. The one thing this relys on though is minimal differences in background from frame to frame. It may take some additional manupulation to get rid of all the background on real images automatically (I used some simple patterns to test it). I was mainly interested in doing this to create race training aids (masters training) so I'm wasn't going to be too picky about background artifacts.

This needs to be refined but here is a general method that I think will work:

thresh=.3; % set difference threshold for background removal
a0=a; % set a0 as reference image where background is retained
b2=imabsdiff(b,a0); % difference image (image processing toolkit function, but easy to duplicate)
imagesc(b2);colorbar % look at difference image to check threshold setting
b(b2<thresh)=0; % set background to 0
imagesc(b);colorbar % look at resulting image
c=a+b; % add first and second images
imagesc(c) % look at resulting co-added image
% next need to creat loop to repeat process for other images in sequence
post #11 of 31
It is certaiinly easier to produce the montage if the images were shot from a fixed camera, with no background change. But the quality may be substantially less, since the individual images of the skier will be so much smaller. This is especially true with standard definition video and, of course, becomes more of a problem the larger the area that must be included in the viewfinder. For a short, one-gate sequence, a fixed camera would often be a good solution. And if you don't care so much about the quality, but just want a basic tool for training, it'll do!

It is always easier to isolate the skier from the background if there is good contrast, and the image is sharp. A clear blue sky or blank white snow in the background is ideal. On the other hand, if the background is entirely featureless, it can be very difficult to register each frame where it belongs.

I look forward to seeing your results!

Best regards,
post #12 of 31
Here is one I made in Matlab. If you have frames that don't overlap, computing the differnce between the two identifies the skier (subject to a little manual tweaking). To "paste" I replace the skier pixels in the background image with just those.

rcahill -- unfortunately, I can't find the scripts I wrote.

Photobucket is blocked here at work, so I hope the photo link in the post is still valid.

(As to the skiing -- I know, more down the hill...)

The post:

direct photo link:
post #13 of 31
Here is the initial step for a different mosaic -- I used the rock to register the frames, before isolating the skier. (I saved the offsets and then went back to the original single frames to get the skier.)
post #14 of 31
Thread Starter 
It may be a little while before I get results. My wife broke her leg practicing running gates on her snowboard Jan 1 for an upcoming Teamski league race out here (rec GS). I'm splitting my time between nursing (requires beer and wine to do effectively), house work (maybe not so much beer and wine), working, and trying to get in some skiing myself. All I have for cameras is a Flip Video Ultra and and old Canon G2 point and shoot digital. (I tested my Matlab montage on some alignment images from a 100 milion dollar space based ozone sensor I'm working on at work though... your tax dollars at work)
post #15 of 31
OK there are a lot of ways to do this...

Here's the down and dirty method, super quick, super easy, super precise way to sequence. You'll need Photoshop (anything above 5.5 should do fine, I use 7).

Take your background image, then select one of your sequence images. Use the lasso to select around what you want. You don't have to precisely outline just take a big chunk. Paste it to the background and it will go in as a new layer.

Line up the paste to the background with a loose approximation. Then change the paste layer blending option from "normal" to "difference" and carefully align the layer until background items which should match up black out. You'll know it when you see it. This method of alignment is very precise, very easy, and very fast. When it is lined up, change layer blending to normal.

Precision adding:
Under the layer menu select "add layer mask, hide all." Now you can select a white paintbrush tool. Select the layer mask of the pasted layer from the layer window. You can now literally paint in that element of your sequence. If you paint in to much, switch to black on your brush and you can paint out parts.

Repeat as necessary for each element of the sequence. This is infinately faster and easier than the simpler method of painstakingly selecting the precise area for copy, visually aligning, then smudging the edges of the paste. You can also turn layers on and off if you decide you don't want an element.

When you are done, save your PSD file. When you want to export you will need to flatten the image from the layer menu. Then you can save as JPG.

post #16 of 31
Thread Starter 
I played around with some Matlab code some more to try and generate a montage automatically from a list of images. The attached image (If I uploaded the url right) is a montage of several frames from my low resolution Flip Video Ultra (480x640 @30pps). I had to add some additional code to deal with a complex background. It ran in 4 seconds. If anyone is interested, here is the code (you need the image processing toolkit):

fd='c:\raydata\m\'[list,filepath]=uigetfiles('*.bmp','',fd); % select images
nf=size(list,2) % sort

se = strel('square',sesize); % create structure for following functions
thresh1=20; % set difference threshold for background removal

a0=imread([filepath,char(list(1))]); % set a0 as reference image where background is retained
c=a0; % Base image with background

for k=2:nf
b=imread([filepath,char(list(k))]); % read each subsequent image
b2=imabsdiff(b,a0); % difference each subsequent image with base image
q=zeros(m,n); % make dummy image with all zeros to construct mask
for k2=1:3
q(b2(:,:,k2)>thresh1)=1; % set ones for all r,g,b pixels above threshold
q=im2bw(q,1); % convert to binary image
q=imfill(q,'holes'); % fill holes im binary image
q = imerode(q,se); % erode binary image mask to remove noise
q = imdilate(q,se); % dilate back to restore mask
q=repmat(q,[1,1,3]); % replicate to create r,g,b image
b(q==0)=0; % set all pixels outside binary mask to zero in sebsequent image
c(q==1)=0; % set base image overlap with next image to 0 (pixels inside mask)
c=imadd(c,b); % add base and next image in sequence

figure;imagesc(c) % look at resulting co-added image

post #17 of 31
Nice work rcahill.
Stupid question time...
1. What is Matlab?
2. How much is it?
post #18 of 31
Thread Starter 
It is a powerful matrix and array math software tool used by engineers and scientists. The base program is $2000 USD and the image processing toolbox is $1000 USD.


You may be able to find an older copy cheap. You can also get an academic version cheap if you are a student or in an academic profession. It may not be exportable though. I have a personal copy of V6.5 and image processing toolbox 1.0. I used these for my montage program. I have newer copies at work. You could also do this in a similiar program called IDL or possibly Mathcad. I work on space based optical sensors so I have access to a lot of powerful tools.
post #19 of 31

I just finished piecing together my first montage. 


I used a free version of VCW VicMan's Photo Editor.


Some of the edges around the skier are a little choppy/messy, but not too bad for my first try I think.


post #20 of 31

Nice work, JPH!


Whoa--just viewed this thread for the first time since the new software platform change. The full-size, instead of resized, images may not be such a great thing after all!. I liked the old software's resizing to make photos fit better on the screen, while allowing you to click on them to bring up a full-sized version in another window.


Alas--growing pains!


Best regards,


post #21 of 31

Bob, nice to hear you are an accomplished photographer as well.



Yea, the images seemed to fit on the screen well, you are corrrect. I hate over sized pictures on forums.

Fixed position is really critical for shots like this. If you can, I'd weight down the tripod too just in case.

post #22 of 31

Hey--my images have all been resized now! Much better, I think. Thanks to the EpicSki Tech Team. Changes are happening quickly around here!




Yes, if you're using a fixed camera, the more fixed the better. A solid tripod with extra weight (I sometimes tie a stuff sack to the tripod and fill it with rocks when I'm carrying a lightweight tripod. It's a good trick!)


But I rarely use a fixed camera for these montages. I like to shoot as full-frame as possible for the highest-quality image, and follow the skier. Fixed camera position works well, though, for brief sequences, such as one single slalom gate or something.


Best regards,


post #23 of 31

Whoa!  Summit, can you tell us who the skiers are in the pictures?

Edited by alexzn - Tue, 03 Feb 09 16:37:04 GMT
post #24 of 31


Originally Posted by alexzn View Post


Whoa!  Summit, can you tell us who the skiers are in the pictures?


Certainly. The skier in both of my photos is my friend Brett. He goes by Blurred Elevens on TGR.

post #25 of 31

Incredible content in this thread. Thanks for sharing know-how. I can't wait to make a montage.


If this thread doesn't scream "wiki", none do. 

post #26 of 31

Thanks, hoody. I thought it had already been turned into a wiki...has it been removed?


Meanwhile, it appears that my images no longer scale to fit, once again. Or is that just my browser? I know they were "fixed" once!


Best regards,


post #27 of 31

Well, here it is--the wiki based on this thread (click here). To date, it includes only my post above. There has been a lot of great information from others since I posted that, so the wiki could stand some updating!


Best regards,


post #28 of 31


Originally Posted by JayPowHound View Post


I just finished piecing together my first montage. 


I used a free version of VCW VicMan's Photo Editor.


Some of the edges around the skier are a little choppy/messy, but not too bad for my first try I think.


Did you have any trouble downloading this software? 

post #29 of 31

Thanks for all the tips. Just finished my first montage. Used images that were shot using a DSLR, at 6 frames per second. Used Pixelmator on a Mac to put the images together. Here's the result:



I had more pictures that i hopen i could have used in this montage. But when shooting these images i was rotating left as i followed the skier with my camera. Because of this and also because my zooming was changing, I could not use those images to extend this montage. But I know better next time how to position myself to shoot better material to be used into these.

post #30 of 31

Would be interesting to get an update from those actively creating montages and see what's changed in the last 4-5 years.


Also, the Wiki link in post #27 is a dead end now.

New Posts  All Forums:Forum Nav: