Autor: Redaktion

Understanding RAW

RAW is an uncompressed video format.

The recording of raw sensor data, so-called RAW data, is common in professional high-end camera systems and is made available for a few more compact camera types as well due to external recorders.

In the following, RAW will be explained and distinguished from terms like log and uncompressed video. All three preceding terms describe technically completely different approaches, but all pursue the goal of storing as much information as possible from the recorded signal. 
The term RAW describes the completely unprocessed data stream as it passes from the sensor to the memory. Raw data are therefore in no way viewable images, but actually only a stream of information about which brightness levels were measured where on the sensor surface. These analog measurements are then converted into a digital representation of this information by the camera-internal A/D converters. Depending on how these A/D converters work, a raw data stream with 12 or 14 bits, for example, is thus generated. If the sensor data is quantified with 12 bits, 4096 brightness gradations can be stored per pixel. 14 bits allow the distinction of 16384 gradations per pixel.
[size=3][font="Times New Roman", serif]If the raw data is saved as a file, the camera creates a container - the format is usually manufacturer-specific. In addition, any camera settings such as white balance, ISO settings, etc. are also saved in this container as metadata. The interpretation of the recorded data according to these camera settings, which would be included in these data before a finished video image would fall out of the camera, is shifted to post-production. This makes it possible, for example, to change color temperature settings later without sacrificing quality.[/font][/size]
Since image sensors cannot "see" any colors in principle, methods always come into play that allow the sensors to see only the primary colors. 3-chip CCDs used beam splitter prisms. However, this method has brought with it some design limitations. Since CMOS sensors have made an enormous step forward in development, most cameras have single-chip CMOS sensors with Bayer patterns. A Bayer pattern is a color filter arrangement on the sensor surface that allows only the red, green or blue light components of an incident image to pass through for each photosite.

RGB Bayer Pattern (Source: GPL, WikiCommons, Cburnett)
The process that creates an image is called DeBayering. The comparison with the development of a film negative is useful here. Raw data can be reinterpreted retrospectively with appropriate software - this enables maximum quality to be obtained from the sensor data. Since you usually work on computers for post-processing that are much more powerful than the camera-internal graphics computation, you can also use much more computation-intensive algorithms to create video images from the raw data. 
The fact that data is recorded as raw data does not necessarily mean uncompressed. Usually RAW is also compressed, just like conventional video data. REDCODE compresses at rates from 3:1 to 18:1. Sony's F65 RAW also has a compression rate of up to 6:1. ARRIRAW , on the other hand, is an uncompressed raw data stream. The degree of compression can have an influence on the image quality, but especially the lower compression rates are considered to be virtually lossless.
Log video is, as its name suggests, video - and therefore not RAW. Camera manufacturers like ARRI, Sony, or Canon have developed their respective log curves to store a maximum of dynamic range in a standard video signal. Log video recordings therefore appear low-contrast and desaturated. Only with appropriate LUTs the images do become pleasantly viewable.

Related Articles: