I've been experimenting with a new technique for making portrait photographs that pop. In situations where I would previously have used light to separate my subject from the background I am looking instead into using depth of field. This is nothing new of course and I used it in my 50 Strangers set but it does have its limitations when shooting wider environmental portraits. Lenses simply aren't fast enough to create a super shallow depth of field at wider angles.
Enter Bokeh Panoramas a technique initially conceived by an amazing wedding photographer called Ryan Brenizer and often referred to as the Brenizer Method. The idea of stitching lots of photos together is not new but what Ryan did was come up with a new application for the technique. The basic idea, shoot with a medium telephoto length lens with a wide aperture, such as an 85mm f1.8 on a full frame body, to get a shallow depth of field and stitch multiple frames together to create a wide angle shot.
Ryan has many great examples over on google+ and a B&H video explanation on his blog. They are well worth checking out. One of the best tutorials I found was here by Daifuku Sensei. The thing I liked about it is that it clearly shows the number of frames required to create the panorama and how they overlap each other.
I had several failed attempts at the technique but now have some of the best pictures of roughly human sized sign posts and statues ever taken! When I first started trying the method I was reluctant for some reason to overlap my shots too much. I was worried that when moving on to a human subject this would segment the person into too many frames and make the stitching problematic. This is still a concern but what I have realised is it's better to have too many frames than not enough and then to apply masking to ensure the best frame of the model is used by the stitching program.
What I have learnt is the key frame, the one including the model's face is the one to get first. Meaning you can worry less about your model moving. But before you take any frames you need to let your model know what you're about to do. Explain that you just need them to hold a pose past the point when they hear the shutter go, to hold it even though you appear to be randomly shooting something else and most importantly not laugh at your impression of James Cordon doing the robot in Gavin & Stacey! If they do move it's not the end of the world, but setting the scene and trying to minimise movement will make your life simpler. You also need a clear idea in your head as to where the boundaries of your shot need to be. Just how much in each direction of your subject do you want to cover? Then come up with a method of ensuring you cover the entire area overlapping each frame by 1/4 to 1/3 of a frame.
Personally I've found holding the camera in landscape orientation seems more intuitive and easier to ensure I've covered the entire scene, but technically it makes no difference. You've still got to overlap in all directions in the same way.
So once you've taken all of your frames, what then? Well the next thing I've found is you need some serious computing power or a love of drinking tea whilst it stitches them together for you! Crunching all of those 1s and 0s is not quick. Photoshop CS5 can automate the process of creating a panorama for you and worked very well on some of my test images. It results in a file with multiple layers, one per frame and then adds a mask to each layer to include only the necessary parts of each frame. I liked this method as it meant the mask of each frame could be tweaked slightly should there be movement between each frame that you wanted to compensate for manually. With more complicated examples including more frames it simply seemed to give up though and did not complete the masking process.
I then found an application called Hugin and best of all it's free! It seemed to work well where Photoshop had failed and produced great results.
Through trial and error I have come up with the following steps for stitching the images together successfully:
- Import the images into Lightroom
- Perform my usual pre editing work flow on the images which includes exposure, brightness, contrast, levels, saturation and curves type tweaks to get the best base images prior to full retouch or other post processing.
- Most importantly ensure that Lightroom's Lens Correction feature is used to remove the vignette and distortion on the raw image. This step doesn't seem to be stressed in other tutorials I have read but I found it essential. I was not happy with Photoshop's ability to remove the vignetting and did not investigate Hugin's built in ability due to step 4 below. If you do not remove the vignette from each frame then the images will not blend effectively. You will clearly see a dark outline around each frame in the final image. Obviously the lens used will effect how much of an issue this really is but my Canon 85mm F1.8 seems to vignette a lot
- Export the images to TIF files at a resolution somewhere in the region of 2000-3000 along the long edge. This is why it is important to do vignette removal in something like Lightroom first so it is removed from your working set of images. These exported TIFs will eventually be deleted. I don't like working with JPG files as they are lossfull and am never sure what quality setting is optimal so working with uncompressed TIFs seems sensible. The 2000-3000 reduction in size of the images is so that your photo stitching program has less to crunch. Unless you've a billboard to hand you're certainly not going to need a full sized panorama so reducing at this stage does not cause an issue and will speed up the processing. Once the images are stitched together you will have more than enough pixels to play with
- Launch Hugin and load the images using the Assistant tab
- Once the images have be loaded, which includes Hugin performing analysis on them, click the Align... button
- The Fast Panaroma Preview will be shown once alignment has been completed. Here you can see how it has stitched your images together on the Preview tab and which images have been included on the layout tab
- If you don't wish part of one of the frames to be included then it is possible to mask out part of an image using the Mask tab in the main window. I used this to remove part of the stick from one of the frames because it had moved between shots and I wanted to ensure the moved version was not included in the final image. This is where Photoshop is better in my opinion as this masking process can be done at leisure at the end. Hugins masking is not as simple as it involves drawing polygons around areas you wish to mask out.
- Click the Create panorama... button on the assistant tab and go and make a cup of tea!
- You can now delete the working set of TIF images and complete any retouching or post processing on the resultant panorama in Photoshop
Once you've created your panorama Brett Maxwell has written a web page which allows you to work out what the equivelant focal length and aperture you would have needed to take your image. Which is quite fun to find out. The image above for example was shot with an 85mm lens @ f1.8 and is mad up of 26 frames resulting in a 149 megapixel image equivelant to 32mm @ f0.68!
subscribe via RSS