How to transform an image's area into a cube - with photos

I am doing a project for a client and I need to manipulate a photograph; either by photoshop or by taking the shot at a specific angel/lens.

My goal is to turn all the quadrants of photos into a cube (6 faces) while constraining to client demand.

Here are the photos aligned as an open faced cube:

http://prntscr.com/avdj9t Now if you already figured it out, this folds and becomes a box.

Another thing that the client wants is to have the edges of every face connect perfectly so in this photo all the faces are interconnected:

This image when rotated, can fit any of it's neighbouring faces, as you can see: http://prntscr.com/avdk45

So essentially, the question is: how do I create a photo like this where the end result would be a folded cube that all its faces interconnect forming a printed cube with a person inside it. Thanks

Sorry for not posting more photos here since stack has a limit of 2 photos until get 10 points (sheesh)

thanks again :)

• Your sample images are the opposite of what you describe you are trying to do. They show a view of what the rest of the world looks like from a single point inside the cube. Your description is of what the inside of the cube would look like from the rest of the world. The problem with that approach is that each face of the cube would look different when viewed straight on than when viewed from an angle so it will be impossible for each panel to be seamless to the adjacent panel. – Michael C Apr 22 '16 at 15:39
• @MichaelClark that's not entirely true. 5 shots perfectly rotated around the no-parallax-point by 90° (ignoring the downward ground view) will distort when viewed off-axis, but their edges will correlate, resulting in a "seamless" transition to the adjacent panel. There would be slope discontinuity at the seams due to sudden perspective distortion change, but perfectly rotated shots will line up pixel for pixel (assuming no motion of the scene such as wind-blown trees, etc.) – scottbb Apr 23 '16 at 19:50
• Only if you are looking outward from the cube and not inward. What the question describes as the goal is a cube with pictures on the faces that appear to look like a display case with a head inside. If you are perpendicular to one of the faces the head should be in the middle of the panel. If you are at a 45º angle to two faces the head should be centered on the seam between them. You can't have both at the same time. – Michael C Apr 23 '16 at 20:39
• Folding the cube inside-out doesn't change the edge discontinuity. From a topological standpoint, the camera's point of view of an inside-out folding is at infinity in all directions. – scottbb Apr 23 '16 at 20:42
• I don't see how you get that from the question at all. Are you saying the camera positions are all aimed at a central point, taken from 6 orthogonal viewpoints? – scottbb Apr 23 '16 at 20:52

The only way to get the cube faces is to cover the cube by shooting a spherical 360ºx180º pano, first. This is non-trivial.

There are cameras that can create this type of image in a single shot (e.g., Ricoh Theta), but the image quality and resolution of these types of cameras tends to be very low and you have no control over the stitch, so panorama stitching is how most professionals do these types of images. But it does tend to take a few months to years to amass both the equipment and shooting/processing skillz to get good at this, and the specialized equipment (fisheye lens, panorama head, stitching software license) is not cheap.

Creating and working with this type of panorama is most typically done in equirectangular projection. This format is most convenient because it contains the entire panorama in a single 2x1 image, and is going to be the format for packages like Pano2VR, or Flexify (commercial Photoshop plug-in) or Hugin to remap to other projections, such as cube faces or origami balloon (i.e., print it out, fold it as an origami balloon, and get your printed cube).

My personal recommendation for your specific situation would be the Flexify plugin for the ease-of-use, the ability to preview the results, and the large number of remappings, which include making a box with tabs to fold and glue, but YMMV as it only does still-image remapping and requires a Photoshop license. Pano2VR is much better if you also need to deliver an interactive pano in HTML5 or Flash.

Similar remapping issue: How are "Little Planet" photos created?

• would there be a sort of emulator for android that you know of that can take that kind of photo? – erezT Apr 24 '16 at 7:49
• @erezT, suggest you ask that as a separate question, and since we don't do product recommendations, possibly on softwarerecs.SE. – inkista Apr 24 '16 at 15:12

So essentially, the question is: how do I create a photo like this where the end result would be a folded cube that all its faces interconnect forming a printed cube with a person inside it.

You're basically creating a spherical image -- one that spans 360° horizontally and vertically. You obviously can't take such an image with a normal camera, so the usual thing to do is to take several photos and stitch them together. There are various pieces of software that will do the stitching; searching with terms like spherical photo stitcher will turn up several.

But you're not really talking about doing the usual thing. In your case, you really just need six square images that align at the seams. Conceptually, you'd want to take six square images (north, south, east, west, up, and down) that each span 90° horizontally and vertically. On a full frame DSLR, an 18mm lens will give you 90° horizontally at a 3x2 aspect ratio, but you'd need to go even wider to get a square image that gives 90° in both directions. So you may need to take more than six photos, or you could cheat a bit and stretch the images vertically.

I can't think of a way to automatically align the images in Photoshop, so there may be some hand work there.

Alternatively, you could create a spherical image using one of the stitching programs above, and then figure out how to extract square images in each direction from the spherical image.

• >koff< 360 horizontally x 180 vertically. :) – inkista Apr 22 '16 at 17:28
• Hey caleb :) So re this line: "On a full frame DSLR, an 18mm lens will give you 90° horizontally at a 3x2 aspect ratio." - I'm currently in development without any gear (nonetheless a 18mm lens) so would there be a sort of emulator for android where I can imitate those specs you gave me? thx – erezT Apr 24 '16 at 7:47

Turn all the quadrants of photos into a cube (6 faces)

Ok That specific thing is a cubical projection. It is very common on 3D rendering and virtual tours.

Here is a little explanation on the diferent projection types: http://www.otake.com.mx/Apuntes/Imagen/EnviromentMaps/

You in theory can make a direct cubical projection but the distorsions are veeeeeery specific, so the best aproach is to make first a sperical projection and then transform it into a cubical one.

In fact on that same page there is an explanation on how to take the photos for a spherical projection (1st step).

I need to manipulate a photograph; either by photoshop or by taking the shot at a specific angel/lens

The second step is to take thoose photos and make a spherical projection. You need a diferent program than Photoshop. Use Hugin http://hugin.sourceforge.net/

The third step is to transform that into a cubical one. Take a look at this How do I make cubemaps with Hugin?

• would there be a sort of emulator for android that you know of that can take that kind of photo? – erezT Apr 24 '16 at 7:49
• Not for cubical. Apps for normal every day users are just for panoramas; a croped cylindrical projection. A spherical one and a cubic one are very specific. – Rafael Apr 24 '16 at 17:38

Stiching multiple image, probably shot on a fisheye lens and remapping using Pano2VR, or Flexify will do the purpose.