Miniatures are a cute imitation of real life. They're fake in a pleasant and palatable way. But what is it about miniature photography that exudes this spurious quality? This project identifies their oversaturated look, scale, and selective focus as the primary reasons. Tilt shift cameras are able to replicate the selective focus naturally with their ability to shift the depth of field in an atypical manner. But this effect can be achieved with selective blurring that will create the illusion that the camera was actually really close to the subject.
A focus line will determine the area in our fake miniature on which the camera is focusing on. A good focus line will highlight objects which would realistically be included in the same depth of field if they were miniatures.
The focus line is set by selecting two points off the original image. The first point will mark the center of the dof. The second point determines how large the dof area will be. By selecting the points on a "per image" basis, different size images can be used without worry.
To complete our fake depth of field, the areas of the image that are not "in focus" must be realistically blurred out. This is accomplished with a Gaussian blur that increases quadratically alongside the area to be blurred out. This means that the Gaussian blur will be convolved with itself every time the blur distance increases (by the blur distance * 2). The initial blur area is determined by the distance of the second point from the first point.
|
|
|
|
|
|
|
|
To emphasize to "toy-like" qualities of the image, we want to adjust the colors to look artificially saturated. The image is converted to an HSV representation. The saturation is easier to manipulate in HSV; it is simply a matter of scaling the saturation values up by 1.5. When converted to RGB again, the colors pop in an unrealistic fashion, even in the "defocused" areas.
|
|
|
|
|
|
|
|
|
Achieving a miniature-like image doesn't exactly rely on the most complex of techniques we've learned in this class. Like Professor Efros says, the most important component of image manipulation is the image itself. As seen in the failed miniature of Lil B, the proper perspective and objects to group in the depth of field are essential to fooling the viewer.