Illegal Colors and RGB vs YHCh

Umm, this is another long-winded post dealing with colors, COLORS, COLORS!  Move along if you are in hurry... Hopefully the BND / CIA / KGB won't shut down this site because of this thread

Seriously, the VIC-20 and TED series (Plus/4 and C-16 in particular) have YHCh colors which are impossible in "standard Red, Green, Blue" color model. You can find color models of VIC-20 here on wikipedia, but I want to focus a bit on how "YHCh" (my term for generic YIQ/YPbPr/YUV) differs from the "de facto" RGB standard of modern PCs (see the "sRGB" article on wikipedia).

To start, I'll show two common color models: RGBW and YHCh.  Note RGBW is just my name for sRGB rendered in polar coordinates, and YHCh is my name for generic YIQ/YPbPr/YUV.  If you know of an official name for this concept, do tell! (and don't say YUV).  Anyway, an important thing about both is that I am showing 5 slices of the color cylinder... each slice shows all values of "chroma" (from 0% [gray=no color] at the center to 100% [max color] = along the rim of circle) and all hues (0 degrees=Red [up/north], 120=Green [right/down], 240=Blue [left/down])... the 5 slices are at "brightness" values of 32, 80, 128, 176, and 224.  The "brightness" is defined differently in each color model and is explained below.


So first is a set of images using sRGB color space... or as I like to think of it, RBGW (Red, Green, Blue, White... because it is the perfect complement to Cyan, Magenta, Yellow, blacK used by most printers).  There are five "brightness" values are based on the center pixel which is has R = G= B = value (where value is 32,80,128,176 or 224 like I said before).  Pixels away from the center deviate from this value in order to generate an appropriate Hue and Chroma (described above).  Importantly, you should note there are many "illegal" colors (shown in gray) when the "brightness" is low or high...



The sRGB color space is a cube, but it has been rotated onto a corner to generate the "cylindrical" coordinate system in the images shown above (I call it RGBW for lack of an official name). I find it intersting that the "cube" nature of RGB is not very obvious in cylindrical coordinates... The middle color wheel (W=128) is probably familiar to people using various image editing applications... this is the center of the "RGB cube". Not so obvious (my opinion) is that the other color wheels which are from the same "RGB cube", but at different values of "Whiteness"... which typically shows up as a triangle (most obvious in the first and last color wheel). In any case, "illegal" values are shown in some shade of gray. There are many illegal values in this "cylindrical interpretation of RGB"...  do comment if you know a good name for this color model !! (and don't say RGB)

In contrast, and perhaps more importantly for CBM software, I present a set of YHCh images (generic YUV/YIQ images) for comparable values of "brightness"... in this case, the "brightness" is the Luma (equivalant to Luminance because there is no gamma) and represents the perceived brightness by a human (according to ITU 601 chromanicities which apply to all CBM machines).  For each color wheel in the series of 5, you should find that all colors in the wheel have the same brightness... which is very easy to verify if your monitor has a "color" or "chorma" (sometimes wrongly called "saturation") that can be tuned to zero.  At zero "color" all pixels in each color wheel below should appear identical.




As you should see, the analog color model ("YHCh" for lack of a better name) that is used by VIC/TED/C16/C64 (and C128 40-colum mode) has many colors that are illegal on your pathetic PC monitor

Seriously, look up at the middle (Y=128) color wheel in the above image; about half of the pixels are "illegal" on an RGB monitor. If you look at the first or last color wheel (Y=32 or Y=224 respectively) you will see that most YIQ/YUV colors are "illegal" on a modern PC. On the other hand, you should note that YHCh (either YIQ or YUV depending on your CBM model) is not "superior" to RGB... this is also important! Some examples:

•In the bottom color wheel (Y=32) you should see that RGB "maximum Blue" is illegal (it gets clipped outside the circle)
•In wheel 4 of 5 (Y=80) you should see that RGB "maximum Red" is illegal (it gets clipped outside the circle)
•In wheel 2 of 5 (Y=176) you should see that RGB "maximum Cyan" is illegal (it gets clipped outside the circle)
•In the top color wheel (Y=224) you should see that RGB "maximum yellow" is illegal (it gets clipped outside the circle)

In other words, neither YHCh (YIQ/YUV) is better than sRGB, or vice-versa... they are complemenetary... they each have their own unique limitations... or "personalities" as you might call them...

OK, ENOUGH WITH THEORY!!

The above was just some color-theory background (please post if you have relevant links or comments) and below is some "real-life" demonstrations of the color differences you might expect from software.

So, where to begin? Well, Timanthes is just insanely complex if all you want to do is convert an image (so no images), and CSAM is so new (to me) that I don't have anything to compare... but I think I can provide some comparisons for you: Mike Kircher + Tokra versus Hydrophilic (me). Each has its own qualities... which is best remains for your decision.

Thus, below are some images, presented 3 at time: the source image (err, twice), the Mike/Tokra (RGB) conversion and my ImageWork (YHCh) conversion...

First in the set is "Terminator 3". The terminator is cool. Both converted images look accpetable to me, although if you look closely you will see some differences. Which is best? Not much difference here...

\

Second is a set from the music video "Toxic".  These are interesting to me... in the RGB (Mike K.) images, the "almost" white wall and phone both (left side of image) renders VERY smoothly (also the bright light at center/far right).  But the flesh tones of the girl (Britney) are very blocky (she seems to suffer from some skin disease).  On the other hand, my YHCh images are VERY blocky in the "almost white" regions; but the flesh tones are quite smooth.  Generally I prefer to emphasize "important" flesh tones over "boring" gray/white backgrounds... but in this particular image, there is a large amount of "almost white" and less "flesh" so overall I think Mike's RGB version is marginally better.  But do note that my colors of YHCh are more "true" than Mike's RGB at least in the right half of the image... I hope you agree!  I do admit that part is very "boring" / not important part of the image (which is why Mike wins by a bit).


However, blockiness / color accuracy is not simply a matter of RGB versus YHCh color model... it really, REALLY, REALLY depends on the image.  So lets have a final comparison... This is from the "impressionist" artist George Seurat and his famous painting "A Sunday Afternoon on the Island of La Grande Jatte" (si vous préférez Français: Un Dimanche après-midi à l'Île de la Grande Jatte)

It seems Mike's RGB conversion is worse (MORE blocky) in this image than my YHCh conversion.  Also I think my colors are more "faithfull" to the source image.  Well, have a look and decide for yourself   Assuming you agree with all my opinions posted (about all the images), then you should agree that neither RGB nor YHCh is superior, in general... but in some images one may appear remarkably better than the other... to spare everyone from Too Much Information, I have not tried (nor would this site allow) me to post hundreds (thousands?) of images I have played with... so you'll have to trust me... they are mostly the same!  I believe this is because of the exteremely limited 4-bit nature of the CBM video chip (VDC in the images shown above).  With more bits, a clear winner might emerge...

Post Script...
Wow you're still reading?  OK, here is some bonus info for you.  All the images above (except the color wheels) were originally in 24-bit (true-color) RBG and were converted into limited (4-bit) color for the 80-column video chip of the C128 (the VDC).  Only Mike/Tokra's software, and my ImageWork allows this!  If you know of any other (public) BMP/GIF/JPG/PNG/PPM that generates a VDC image, please tell!!!!!!  (Top secret: I have been forbidden from posting info about Dirk Vrooman's VDC image converter.)  CSAM and Timanthes only work on C64 (or C128 40-column).  And nobody else works on TED chips (Plus/4 or C16)... well not that I know, but please tell if you know otherwise!  In short, my ImageWork is the only known software that supports all chips (VIC-I, VIC-II, TED, VDC).  If you know of some other image converter that supports them all, then please clue us in... thanks!!