RED One Image Processing

(4.2)   RED Camera Manual User Help Guide  >  RED One Theory of Operation  >  RED One Image Processing

Image Processing

You're probably saying ,"Didn't we just talk about this?"  The answer is basically yes, but this next section restates some of what we've talked about a little differently, and adds another few layers to the picture.

Let's tackle it piece by piece.

Pixel defects are at the sensor level, no sensor can be 100% perfect all the time.  So pixel correction attemps to correct for any black pixels on the sensor.  Also, there are some wierd patterns created when puttng those red, green and blue channels next to each other.  Nothing we need to worry about. - happens automatically inside the camera.  And we know what 12-bit data means from the previous chapters.  RAW means that it is not processed.

To tackle this one in depth is a little beyond the scope of this discussion - but we're going to give it some of the time it deserves.  This is the first time we've introduced the idea of "codecs" and compression.  A codec is a defined standard for the COmpression and DECompressioin of data.  This means we get an image on one side, we compress it for transmission either over the air or, in the case of the RED by e-SATA cable, then we decompress it on the other end for viewing.  There are many different codecs out there for different uses.  There is H.264 - Apple's Quicktime. Or DVC-PRO is another codec. HDV is another codec.  There are many.  There are codecs for film work and there are codecs for broadcast.  Traditionally the two have not overlapped because of the differences in quality demanded by film.  Enter REDCODE RAW.  REDCODE successfully compresses 4k images at 12 bits of color down to 220Mbits/second.  For comparison, the Dalsa digital film camera has a slightly lower resolution but more colors at 16 bits of color, but its bandwidth requirements are approximately 3.2Gbits per second or 14.5x as much as the RED.

 Another important aspect of codecs - there are destructive and non-destructive codecs.  Some examples of non-destructive codecs  are "Animation" or "Uncompressed 8-Bit".  These codecs store the information pixel for pixel.  They are not used for HD or above generally because they require huge amounts of bandwidth because they are un-compressed. 

A second type of codec is a destructive codec.  Jpeg and GIF are examples of destructive codecs.  JPEG more than GIF, but in both cases, colors that are outside of a given color space are removed.  Basically you're left with an image that looks the same to the eye, but doesn't have nearly as much color information as the original.  "How do they do that?" is the magic question for us.  Because if they are doing it in a way that degrades the original image - then its not acceptable.  The second half of the above quote tells us the answers to some of this question.

REDCODE is a destructive codec.  It does in fact, throw away some of the color information - but because its logarithmic - it can get rid of information far outside the needed range.  At least that's the idea.  And finally, "wavelet based," describes another algorithm for compression over time.  Too complicated or boring to discuss here. If we believe the RED data - which I do - then what comes out on the other end is "RAW 12bit color at up to 4K resolution."