Digital compositing

src: patchinpictures.com

Digital compositing is the process of collecting multiple images digitally to create final drawings, especially for prints, moving pictures, or screen views. It is a digital analog composite optical film.

Video Digital compositing

Math

The basic operation used in digital compositing is known as 'alpha mixing', where the opacity value '', is used to control the proportion of two input pixel values â€‹â€‹that produce an output pixel.

As a simple example, suppose two images of the same size are available and they must be compiled. The input image is referred to as the foreground image and background image. Each image consists of the same number of pixels. Compositing is done by integrating information mathematically from the corresponding pixels of two input images and recording the result in the third image, called the composite image.

Consider three pixels;

foreground pixels, f
background pixels, b
combined pixels, c

and

?, the opacity value of the foreground pixels. (? = 1 for foregrounded background,? = 0 for a completely transparent foreground). A monochrome raster image where the pixel value is interpreted as an alpha value is known as matte.

Then, considering the three color channels, and assuming that the color channels are expressed in = 1 color space (that is, the measured value is proportional to the light intensity), we have:

c _r =? f _r (1 -?) b _r

c _g =? f _g (1 -?) b _g

c _b =? f _b (1 -?) b _b

Note that if the operation is done in the color space where? not equal to 1 then the operation would cause a non-linear effect that could potentially be seen as an aliasing of artifacts (or 'jaggies') along the sharp edges in matte. In general, nonlinear compositions can have effects such as "halos" around adjacent objects, because the influence of alpha channels is not linear. It is possible for composite artists to compensate for the composing effect in non-linear spaces.

Doing alpha mixing is an expensive operation when performed on all 3D images or scenes. If this operation has to be done in real-time video games, there are easy tricks to improve performance.

c _exit =? f _in (1 -?) b _in

c _exit =? f _in b _in -? b _in

c _exit = b _in? (f _in - b _in)

By simply rewriting a mathematical expression one can save 50% of the required multiplication.

Properties of algebra

When many partial transparent layers need to be compiled together, it's good to consider the algebraic nature of the compositing the carrier uses. In particular, associations and commutatives determine when repeated calculations can or can not be avoided.

Consider the case when we have four layers to blend to produce the final image: F = A * (B * (C * D)) where A, B, C, D are partial transparent image layers and "*" indicates joint operators left layer above the right layer). If only the C layer is changed, we must find a way to avoid re-mixing all the layers when calculating F. Without special consideration, four full-blown images need to be done. For commodity combined operators, such as mixing additives, it is safe to reset the mixing operation. In this case, we may compute T = A * (B * D) only once and merely integrate T * C to produce F, one operation. Unfortunately, most operators are not commutative. However, many are associative, suggesting it is safe to re-group operations to F = (A * B) * (C * D), ie without changing the order. In this case we can calculate S: = A * B once and save this result. To form F with an associative operator, we only need to perform two additional composing operations to integrate the new S layer, by computing F: = S * (C * D). Note that this expression shows compositing C with all the layers below it in one step and then mixing all the layers above it with the previous result to produce the final image in the second step.

If all layers of images change regularly but a large number of layers still need to be compiled (as in distributed rendering), the commutativity of compositing operators can still be exploited to speed up calculations through parallelism even when there is no advantage of pre-computation. Again, consider the F = A * (B * (C * D) image). Each compositing operation in this expression depends on the next one, which leads to serial computing. However, associativity may allow us to rewrite F = (A * B) * (C * D) where there are clearly two operations that are independent of each other that can be run in parallel. In general, we can construct coupling-wise composite operating trees with logarithmic height in the number of layers.

Maps Digital compositing

Software

The most historically significant nonlinear composite system is Cineon, which operates in a logarithmic color space, which more closely mimics the natural emulsion film's light response (the Cineon system, made by Kodak, is no longer in production). Due to the limitations of processing and memory speeds, compositing artists usually lack the luxury of having a system of making medium to linear conversion chamber for compositional steps. Over time, limitations become less significant, and now most compositing is done in linear color space, even in cases where the source image is in a logarithmic color space.

Compositing often also includes scaling, retouching and color correction of images.

Node-based and layer-based compositing

There are two radically different digital composing workflows: node-based composites and layer-based compositing.

Node-based compositing represents the entire composite as tree graphs, connecting media objects and effects in procedural maps, intuitively exposes the progression from source input to final output, and in fact the way all composite applications internally handle composites. This type of compositing interface allows great flexibility, including the ability to modify the parameters of the previous image processing step "in context" (when viewing the final composite). Node-based compositing packages often handle keyframing and bad timing effects, because their workflow does not originate directly from the timeline, as does the layer-based compositing package. The software that incorporates a node-based interface includes Natron, Apple Shake, Blender, Blackmagic Fusion, and The Foundry's Nuke.

Layer compositions represent each media object in a composite as a separate layer in a timeline, each with its own timing, effects, and main frames. All layers are stacked, one on top of the next, in desirable order; and the bottom layer is usually given as a base in the resulting image, with each of the higher layers progressively applied over the previously composed layer, moving up until all the layers have been rendered into the final composite. Layer-based compositing is particularly suitable for fast and limited 2D effects such as motion graphics, but becomes awkward for more complex compositions involving multiple layers. The partial solution for this is the ability of some programs to see the composite element composition (such as images, effects, or other attributes) with a visual diagram called flowchart to nest composition, or "comps," directly to other compositions, thereby adding to the complexity of render-order by the first composite layer in the starting composition, then combining the resulting image with a layered image of the composition being processed, and so on. This example is in the Adobe After Effects program.

src: www.artofvfx.com

Digital compositing

Jumat, 15 Juni 2018