i was asked to explain how i create 3D images from movies scenes.

well, when i'm watching a movie, i always keep pen and paper at hand to write down the timestamps of scenes where i spy the opportunity of creating 3D images, that is, scenes where the camera moves across a relatively still subject, left to right or right to left.

on my full hd VAIO laptop, i make screen prints of two separate frames from such a scene, where there is lateral camera movement (i.e. left to right or vice versa) across the subject, either in a straight line or better still in a more or less circular line around the subject, so that the subject stays in the same spot within the frame. in an ideal situation the subject would be perfectly still (i.e. not moving at all), plus ideally the camera movement would occur in a perfectly horizontal plane and the lens would not give any distortion due to the movement, but you'll hardly ever find that in any movie. if you take two frames that are pretty close you'll increase safety as far as unwanted subject movement is concerned, but you'll often need more distance between shot 1 and shot 2 to create a more dramatic 3D effect, especially if there was a considerable distance between camera and subject(s), but the latter will also raise the challenge level for the viewer.

what i do next, is, i take a selection from the first frame and paste it to the left half of a new photoshop document that is once the height and twice the width of the selection and then copy/paste a second one of the same size from the other image to the right, if the camera movement was left-to-right, or vice versa (if you prefer to work cross-eyed, just do the opposite). i make sure the subjects in left and right are in about the same spot, so if that particular bird is in the top left corner, i make sure it's in the top left corner of both images. especially if your movie was in letterbox format (2.35:1) the size of your parallel 3D image will be pretty limited unless you size down its width, that's why most of my 3D movie pics are more like 16:9. in cross eyed or x-view 3D your limitation isn't that extreme, but you'll find that crossing your eyes reduces the subjective size of the image and perspective and depth are really compromised in cross view as well. this is why i still prefer parallel by a mile.

i now draw a horizontal line from an object or point of recognition in the left image to its counterpart in the right image, to see if they are level more or less. i repeat this for several objects or points of recognition through the whole picture. i do the same for the vertical position of object in both images. if there is a considerable amount of distortion in one frame relative to the other, i place one of the two images on top of the other and switch the top layer on and off and/or i change its opacity, so i can compare top layer to bottom and resize, rotate and skew/distort the top image until it fits the one below.

if you now try and find the 3D image (by either crossing your eyes or staring through the picture), you'll notice that parts that moved between shot one and shot two are unsteady with a flickery quality to it and hard to focus on, almost like ghost images. these parts i copy-paste from one image into the other and to prevent that part of the image from looking like a flat piece of cardboard, i copy/drag separate parts of it towards its counterpart or away from it, one pixel displacement at a time, to create depth (drawing a selection closer to the opposite side brings it to the front and vice versa - in cross view this is the opposite). in photoshop you'll be needing a nice feathering on your selection tool to avoid annoying sharp edges.

well that's about it. hope you found my instructions informative.

cheers, aida.