Archive for category RED
FLUT Color Science from Red One explained
A Look Up Table, is a table of values that represent a mathematical formula that has already been calculated for every value you could want. So instead of doing the calculation you look up the answer you want by tracing along a table. We use look up tables in programming when we it’s easier just to hand over the values than to try to describe the formula and/or when we need speed and have plenty of memory and memory access is faster than the actual calculation. (In some cases the table is combined with calculation)
From the point of view of a user a Look Up Table, just means a transformation, in our case the transformation of one color value to another.
A floating point number is a way of writing a number where the decimal point can move. (In computers its the binary point.) This is incontrast to a “fixed point” number where there is a certain fixed number of digits on either side of the point. Floating points are useful because they allows you to represent really large numbers by moving the point all the way to the end or tiny numbers (between 1 and zero) by moving the decimal point all the way to the front and to use either type of number together in whatever math you are doing.
From the point of view of the user floating points mean, more precision, so the color transformations are less likely to cause posterization for example.
Since floating point numbers are already in use internally in most post production software this could also help make Red’s color science more portable.
by
IBloom
http://www.crimsonworkflow.com/home.htm
Red One Build 20 Test by Art Adams
Given below is build 20 test done by Art Adams (www.artadams.net) for Red One Digital Cinema Camera.
For complete test results you can visit ProVideo Coalition.com (http://provideocoalition.com/index.php/aadams/story/red_build_20_torture_tests/P1/)
Red Raw: Part 2
Color Space:
Color space, put simply, is a mathematical description of color. It is used to represent color in ways that are appropriate to the display
device being used, and to account for the range of colors that the particular device is capable of achieving (the “color gamut”). When
working with Red images, Red has predefined a few color spaces for use when converting RAW images to RGB images. These color space
models are basically intended to provide color values in the converted image that are appropriate for various display methods, based on the general characteristics of those displays (in particular their white point), and their specific color gamut. Essentially, the values that are the result of the debayering process are further altered by each of the color space settings (in both value and saturation) to achieve more
accuracy to the “real world” colors that were present when the images were captured by the sensor, based on the characteristics of the
intended display system. At this point, the color space choices (they are actually color matrixes, and can be described by either term), and their effects on the image, that are currently offered by Red are:
Camera RGB: This matrix passes the RGB values as described above and does not modify them based on any particular display technology.
Rec709: This matrix alters the resulting RGB values to, in theory, properly display an accurate representation on an HD video display. It also seems to add quite a bit of saturation, and because of that, is rarely used.
Redspace: This matrix appears to be similar to Camera RGB, but with a mild saturation boost
At this point, the best choices for most projects seem to be Camera RGB and Redspace, with Red Space / Camera RGB will be the preferred color space for many people including major VFX post house
Gamma:
Like most electronic sensors, the Mysterium sensor represents the world in what we would call Linear Light. In a linear light representation, absolute values are obtained based on the brightness of the elements of the image.
Each chip is twice as bright as the previous one. However, human eyes do not perceive this accurately. For instance, in theory, the brightest grey chip is actually 64 times as bright as the darkest chip – yet it doesn’t “look” that way. Additionally, if a linear light image is observed directly on an electronic display, it will look much too dark, because electronic displays are not designed to display linear light. Displays have a gamma characteristic – in other words, they are nonlinear, particularly in the darkest areas. There are a number of technical reasons for this, but suffice it to say that in order for an image to display properly, it must have the same gamma characteristic as the display being used. In order to make the image more “perceptually linear” on a specific type of display (one might call this “linear luminance” rather than “linear light”), the values are altered by a luminance curve that effectively boosts the lower values in a nonlinear way. The value of this curve is the gamma, and in most cases, it is designed to match the gamma characteristic of the display system. Another option is to encode the linear light values to logarithmic values, which has the effect of redistributing the available value levels. providing more available values at the lower and middle end of the scale, and fewer at the high end - much the way human eyes perceive light. By making the lower values more precise, and the higher values less precise, it also allows you to fit more useful information into fewer bits, allowing, for instance, the 12 bit linear light information provided by the sensor to comfortably be represented in a 10 bit file. And finally, it has the advantage of representing the scene in values that are more like negative film, which also has a logarithmic response. It must be noted, however, that a logarithmically coded image will not look correct on most electronic displays without the use of a lookup table (discussed below).
Red provides different gamma settings that represent all of these possibilities. They are:
Linear Light: A direct representation of the RGB image without any gamma or log encoding applied. This requires more bits – at least 12 bits to maintain the precision of the original values, and preferably 16 bits to maintain the precision of the demosaic mathematical processes. Although the linear light representation is not appropriate for direct display on most electronic displays, it is often desirable for visual effects work due to its mathematical precision and its direct relation to real world physics.
Rec709: A gamma that represents the SMPTE Rec.709 specification is applied to the linear light image. Although commonly referred to as a 2.2 gamma, Rec.709 is actually a specifically defined gamma curve that is similar to, but not exactly the same, as 2.2.
Redspace: Contains many of the characteristics of Rec709, but with a constrast boost, particularly on the higher end, that better simulates a “final” color grade, and is often better suited for on set display.
PDLog685 and PDLog985: The image is encoded to a logarithmic curve that is designed to mimic the Cineon curve, based on film density.
Redlog: The image is encoded to a logarithmic curve that is much milder than the Cineon curve, and better represents the linear values obtained from the sensor, designed to be stored in a 10 bit file. The entire range of 10 bit values from 0 through 1023 is used. The curve is designed so that the precision of the darkest 8 stops of information is retained, and although there is some loss of precision in the upper 4 stops, it is not significant. This setting is often used for transcodes for finishing work, as it best represents the original information with minimal loss, and can be used directly for “video” style color grading without requiring the use of a LUT.
Next post will be working on Log vs Linear.
Excerpts from Assimilate White paper written by Mike Most (Colorist/Technologist)
Red Raw: Part 1
The Red Color Workflow:
The Red camera represents a new approach to motion image capture that relies on post processing, rather than in-camera processing, to deliver images that are appropriate for use in both video and film finishing formats. The files that are written directly by the camera
represent – in compressed form – the actual data captured by the Mysterium sensor, without any manipulation for specific display systems. Because of this, when working with Red images it is necessary to understand some of the basic principles of working with digital images in order to determine the proper path for an intended delivery format.
Without going into unnecessary details, the Mysterium sensor in the Red cameras is a single sensor that captures color images by means of a color filter array that is superimposed on the individual pixel sites on the sensor. The particular pattern that is used is called a Bayer pattern, and consists of alternating rows, each of which has a combination of either red and green pixels (i.e., GRGR etc.), or blue and green pixels (i.e., BGBG etc.). In order to create a “normal” RGB image, a process called a Debayer (also referred to as “demosaic”) is invoked, which uses some rather complex math to predict what each of the individual pixel sites would contain in all three color components by combining the values of the surrounding pixels for each of the colors not directly represented by each individual pixel. A good debayering algorithm can be very accurate, and the algorithm used by Red is very good indeed. Since the red, green, and blue filters used on the sensor are not absolutely “pure,” and the tiny lenses that focus each filter’s light on the image element itself are not perfect, there is a certain amount of “crosstalk” that occurs between the values of each pixel – in other words, the red pixels also contain a certain amount of blue and green, the blue pixels contain a certain amount of red and green, and the green pixels contain a certain amount of red and blue. The debayering algorithm is specifically designed by Red to account for these variations, based on their specific sensor characteristics. The sensor also has a “native” white point, that is, a specific point in the color spectrum that is considered to be white. In order to achieve a more “pure” – and thus accurate to the actual scene – image on a specific type of display, a color matrix is then used that alters each component by adding or subtracting a bit of the other two components, based on the characteristics of the intended display, in particular, the display’s specific white point. In the world of Red, this color matrix is usually identified using the term Color Space.
Will see more about Color Space, Gamma etc in coming posts…
Excerpts from Assimilate White paper by Mike Most (Colorist/Technologist)