A UNIQUE METHOD OF CALCULATING KEY VALUES
For generating the composite image, the matte signal is required to have the blue screen color component at 0% and the foreground object color components at 100%(Figure 20, Figure 21). This process is needed for both digital processing using computer images and also for analog processing using camera derived video images. The matte signal is sometimes called the "key" for the video switcher. "Soft-key" processing can also be adopted to facilitate compositing difficult blue-screened objects such as smoke, hair, or steam, by applying in-between values between 0% and 100%.
The color component of each pixel in the output image will be equivalent to that of the corresponding pixels in the chroma-keyer matte layer plus the pixels in the background image. So, to achieve the best composite image, the color components of chromakeyed foreground imagery should be used for the pixels with a matte signal of 100%, and the color component of background image should be used for the pixels with a matte signal of 0%. For the soft-key portions, the color component of foreground object and background image will be mixed, based on the matte layer values.
Figure: 20 Bluescreen Image (back)
Figure: 21 Mask Signal (back)
Now in this section, we will look at the definition of the matte signal. The color component has three primary attributes: Hue, Saturation, and Luminance. The mixture of Hue and Saturation is called Chromaticity (please see Figure 2). Chromaticity sees no difference between light red and dark red since there is no luminance component. The color cube(Figure 3) is represented by the chromaticity plane with a vertical line crossing on the white point. This vertical line represents the luminance, and every color must be plotted and defined within this cube. The conventional chromakey will be processed within the chromaticity plane, regardless of the luminance. A vector-scope, used for viewing and adjusting the video image, also displays the hue and saturation. Figure 4 shows the key chart created by drawing the contour lines which divide the line connecting the blue point and white point equally. This key chart is used for defining the matte signal.
Figure 22 shows an example of defocused red and magenta objects in the foreground and Figure 23 shows the distribution of color components for every pixel included in the chromaticity plane. As you can see in Figure 23, the belts of distribution extend from the Blue corner to the Red and Magenta corners. The belts between the corners correspond to the edge portions between the foreground and background.
Figure: 2 Chromaticity plane (back)
Figure: 3 Color cube and chromaticity plane (back)
Figure: 4 Conventional keychart w/contour lines (back)
Back to defocused reds. Back to picture quality. Figure: 22 Example of difficult chromakey processing (back)
Figure: 23 Color distribution of a defocused image (back)
If a higher priority is given to keying-out the red object clearly, the key value for the red color component should be set to 100 for the blue-background should be set to 0%(Figure 24). Consequently, key value of magenta gets close to 0%, and magenta object will reflect the influence of the blue-background(Figure 25).
If a higher priority is given to keying-out the magenta object clearly, the key value of edge portions of the red color should be set to 100%(Figure 26) and the resulting image will be as Figure 27 indicates. These limits were necessarily accomplished by using the conventional chromakeying method.
As explained above, contour lines drawn as the key chart will vary quite a bit depending on the color of the foreground object. Contour surfaces utilized in Primatte are very flexible and can be modified to adjust to every color component of the foreground object. (Please see below about the Primatte flexibility.)
Figure: 24 Adjusted keychart with higher priority to Red (back)
Figure: 25 Adjusted sample image using the keychart of Fig.9 (back)
Figure: 26 Adjusted keychart with higher priority to Magenta (back)
Figure: 27 Adjusted sample image using the keychart of Fig.11 (back)
Figure 28 shows an example of chroma keyer material with a Cyan object in front of a blue background. Since Cyan is comprised of color components very similar to the blue in the background, the color differential between Cyan and Blue is very narrow. As a result, steps or ramps of color will show up within the defocused Cyan portions because of the difficulty in drawing contour lines clearly(Figure 29). Primatte, on the other hand, deals with the color in the color cube and takes into account the luminance element. Therefore, it can effectively identify the color difference of even very similar colors.
Back to Chroma Keyer. Back to color difference. Figure: 28 Example of Cyan objects as chromakey material (back)
Figure: 29 Sample image achieved by conventional processing (back)
What this means is that Primatte can flexibly and exactly create key charts by effectively and accurately identifying the color difference between the foreground object and the blue background. A clean matte signal can even be achieved for defocused objects like the ones seen in Figure 22. Figure 30 shows the composite image obtained by using Primatte. As you can easily see, the image quality of the final composited image is far better than the image shown in Figure 22. Primatte is able to realize a high quality composite image by enabling the creation of a flexible key chart accurately depicting the cross section of the color cube.
Figure 5 shows the color cube with the luminance information added. Primatte created this sophisticated three-dimensional contoured surface by defining 128-faces polyhedrons and was able to identify even the smallest color differentials that existed between the foreground object and the blue background. Figure 6 shows the contour surface located in the outer-most space, which describes key differences of 99% to 100%. With another element of luminance added, the foreground object with the smallest color difference(Figure 28) can be keyed-out smoothly and precisely (please see Figure 31 and Figure 32).
Figure: 30 Sample image achieved by Primatte (back)
Figure: 5 Three-dimentional color distribution of chromakeyer material (back)
Figure: 6 Example of contour surfaces created by Polyhedron Slicing Algorithm (back)
Figure: 31 Smooth matte image (back)
Figure: 32 Sample image created by Primatte (back)
Primatte uses three different polyhedrons for controlling the sophisticated three-dimensional key chart and classifies chromakeyer components into the following four categories when processing them:
- Complete background component
- Soft key component
- Blue spill component
- Complete foreground component
As we explained in the previous examples, unique key calculation is realized by processing color data in a three-dimension environment (remembering that Primatte can recognize the luminance!) and generating flexible key charts corresponding to any color components contained in the image.
For the examples in this White Paper, blue background materials have been used for explaining the Primatte chromakey functions. Please note that Primatte can use any color as a background color. Green, red, white, black or even gray can be used.