Building a Color Display Based Upon What the Human Vision System (HVS) Can See

There has been some confusion about how the human vision system resolves image information.  To understand this properly, one has to think about luminance information separately from chrominance information as our eyes handle this data very differently.  Shown in Figure 1 is a 3D plot of luminance contrast sensitivity showing that the typical human vision system (HVS) loses all sensitivity at about 30 cycles/degree in both horizontal and vertical directions  {note: this varies between individuals}.  30 cycles/degree means that over one degree of arc  that typical human vision can resolve and detect as many as 30 line pairs of black and white lines.  This corresponds to a 4.0-inch qHD (960 x 540) display viewed at 32 cm distance.  This same data is represented on a 2D isoplot as Figure 2 below.  Here you can also notice that we actually resolve a little less along the diagonal.

Figure 1 - 3D Luminance Sensitivity Plot for Normal Vision

Now let’s consider chrominance information in Figure 3.  You will see two other lines representing the maximum resolution of 8 cycles/deg  that can be perceived for colors along the red/green line  and 4 cycles/degree along the blue/yellow line (rolling off to zero by 14 cy/degree and 10 cy/degree respectively).  The conclusion from this data is that chromatic data does not need to be so well localized as does luminance data and the vision systems will not be able to tell.


Figure 2 - 2D Luminance Sensitivity Plot


Figure 3 - 2D Plot of Normal Sensitivity to Spatial Frequence Based Upon Chrominance











Now let’s talk about how to build a high resolution display from scratch using what we know about human vision.  A first step could be to begin with a simple black and white display as shown in Figure 4.  In this case there is one subpixel per pixel and you would be able to turn on and off any pixel to achieve any gray level, but with no chrominance information.  For now, let’s assume that this display has a dpi corresponding  to 30 cy/degree.

Figure 4 - A simple black and white display

As a next step, we could take every other pixel and replace that white pixel with 3 subpixels of red, green, and blue as shown in Figure 5.  In this case the luminance of the white subpixel is designed to equal the sum of the luminance of the R+G+B subpixels.  If you turned these subpixels fully on you would see a white screen.  This display still has chromatic resolution at the same 30 cycles per degree in the horizontal and vertical, but the resolution in the diagonal, the direction that has the lowest sensitivity, is reduced by half.  Given that the HVS can only resolve colors at less than half the resolution of black and white, with a display pitch of 30 cycles/degree of pixel pitch in the horizontal and vertical.  On the diagonal, adjacent lines are actually spaced closer than on the horizontal and vertical.  Specifically the line pitch is increased by the square root of two, or 42 cycles/degree , your color data would be at half that, at 21 cycles/degree, in the diagonal direction,  which  is more than sufficient.  Unfortunately, we have higher pitch in a direction where the HVS is less sensitive, so this excessive resolution.

Overall, you would have everything you needed to represent everything that the HVS could see in color, even though you have color data at every other pixel.  Of course, if you tried to accomplish this with a chrominance pitch that was less than 8 cycles/degree for red/green or 4 cycles/degree for blue/yellow, you would begin to see a decrease in chrominance resolution.

Figure 5 - A simple, but adequate, color display

The physical reality is that white (clear) subpixels are far more transmissive than color subpixels since the color filters absorb at least 3X the light per unit area.  To allow the luminance to match that of the R+G+B we must shrink the area of the white subpixel.  If we shrink the width of the white subpixel to one-half the width, and increase the width of the R, G, and B, to be the same width, we would find that the luminance of the combined R+G+B subpixels would be very close to the luminance of that W.  Not only does this balance the luminance of R+G+B to W, but it is far easier to manufacture.

Figure 6 - PenTile RGBW display

The format in Figure 6  is the RGBW that is used by PenTile.  The advantage of this format is that it allows one-third of the subpixels to be eliminated, reducing the number of column drivers, increasing the open area (aperture ratio) of each subpixel and adds a white (clear) subpixel to let even more light through the panel.  Overall, this has the benefit of almost doubling the light throughput of an LCD display. It does all of this without compromising our ability to reduce the chrominance resolution below that which can be perceived.

So, if there is more than enough data for luminance resolution and more than enough for chrominance resolution, some might ask why is it that we can see screen door effects on fully saturated green on black or frailty of fully saturated red or green, single stroke diagonal lines on black?  The data supports the fact that we have represented the luminance perfectly at every pixel and the chrominance to the level that the HVS can perceive, yet we can still see something different than for RGB stripe.

This is known as pattern visibility.  Visibility of the pattern is carried on the luminance channel, so we have sensitivity of this to 30 cycles/degree.  When a block of solid bright saturated color, such as green, is displayed, a checkerboard of green and black, representing lines and spaces, in both diagonals, is visible in the luminance channel at 21 cycles per degree.

Consider also that a single stroke diagonal line of a saturated color.  Since in this case we are not using the W subpixel, this is reduces the cycles/degree by a factor of square root of 2 due to the geometry, or about 21 cycles/degree in this example.  If that diagonal is fully saturated green on black we will have cut the luminous channel cycles/degree by one half since we are not lighting up any other subpixel, so it has the appearance of 21 cycles/degree which is well within our sensitivity for detection.  For this reason it is important to design PenTile displays only for sufficiently high resolution.  Such pattern visibility will be detected at 33% higher dpi than would be the case for RGB stripe.

Just as with half tone printing if you were to fabricate this at too low of a dpi or were to look at this with an eye loop, the HVS cannot blend together the subpixels to cause us to see them as pixels.  Instead we see the subpixels and the pattern that was used to form them.  Done correctly the dpi of the color should be at least 30cycles/degree for us to fully fuse the colors and to be unaware of their exact locus.  Looking from a very close distance or with a magnifier violates the premise of the design to keep these subpixels at levels of 30 cycles/degree or higher.

We are only now at the bottom end of the utility range for PenTile.  This is a technology that anticipates the need for still higher resolution and the requirement for higher efficiency when we do so.  As the industry goes to 300 dpi screens for phones and tablets we are well in the range for PenTile to add more value and to have little or no detectable artifacts.


Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>