Wiggle 3D Displays of Weather Data

Weather is a time varying 3 dimensional phenomena, but the displays of weather data are inherently 2 dimensional.  Since weather phenomena at one level can impact the weather at another level, it would be desirable to view weather on a 3D volume type of display.  Initially weather displays were on paper and currently weather information is displayed on computer screens.   The human mind can interpret 3D volume information in a number of ways.  Having two eyes, the brain can detect the parallax differences between the images for the left eye being slightly different from the right eye.  Objects that are closer to the observer have a greater parallax shift between the two eyes than distant objects.  Hence 3D displays can be generated by generating separate displays for each eye.  The two displays can be directed to each eye using colored filters (such as anaglyph red/cyan glasses), cross polarization filters (such as RealD 3D movie glasses), or electronic glasses which switch the lens on/off in sync with the display (such as 3D TV displays).  While these displays are effective, they require the user to use special glasses with dedicated 3D terminals.  They are generally not amenable to general purpose displays such as cell phones, tablets, or ordinary computer monitors.  These displays also require that the user have two equally good eyes. 

A second way the brain determines distance is through relative motions as the viewer changes the viewing location.  Closer objects have a larger relative motion while distant objects have little relative motion.  This technique is frequently used in movies where the camera is moved (panned) across the scene, allowing the viewer to discern distances.  This is the principal behind “wiggle 3D”.  The Wikipedia article available at https://en.wikipedia.org/wiki/Wiggle_stereoscopy contains additional information on the technique.   Other examples of wiggle 3D images can be obtained using Google for “wiggle 3D”. 

To generate wiggle 3D images, one needs at least 2 images which contain parallax.  Mosher (2012) demonstrated the generation of 3D satellite images using the infrared image to obtain cloud height and then computing 2 parallax images for the visible channel.  The parallax is generated by shifting the pixels left or right an amount proportional to the computed cloud height.  The amount of parallax shift is similar to that which would be generated from the real parallax between GOES-east and GOES-west.  These 3D satellite images are routinely available at http://wx.erau.edu/erau_sat/ for users with anaglyph red/cyan glasses. 

The parallax generation code can also be used to generate wiggle 3D satellite images.  In the development of the wiggle displays it was determined that the parallax shift used in the anaglyph images was too large for a wiggle display.  A reduced shift 1/3 of the anaglyph shift has been used for the wiggle display.  This gives about 8 levels of height in the satellite display.  Using the suggestion from the Wikipedia article, a third image has been added to the display.  Rather than using just the two left and right parallax images, a third image of the original projection also used to reduce the jumpiness of the display. The three images are saved as a single animated gif image.  For a time animation, a JavaScript program is used to loop the animated gifs.  The wiggle satellite images are available at http://wx.erau.edu/erau_sat/for the US areas of coverage for GOES-east and GOES-west.

With the success of the wiggle 3D displays of satellite data, an initial approach at displaying model data in 3D was attempted.  Contours of model heights were displayed for four levels, with the left/right shifts being proportional to the pressure levels.   However, the resultant display just looked like lines jumping around without any sense of depth.   Apparently the brain’s depth perception is based on detection of identifiable objects and lines don’t meet the brain’s criteria for an object.  Experiments were done using various line displays, and increasing the line width with height in addition to increasing the left/right shift does provide some measure of depth perception with a wiggle display.  However having fat lines only worked for relatively smooth fields such as geopotential heights and temperatures.  Data with a lot of variations, such as vorticity or divergence, did not display well with fat lines.  Experiments were done converting the model grids into images.  However the images are solid, so to see the levels below the top, holes in the images are required.  But holes destroy information.  Initial experiments with holes first remapped the initial model image into a higher resolution projection, and then took out random holes.  Having more holes in the upper levels than the lower allowed the five levels to be recombined back into a single image.  While the original information from various levels could be seen with enhancements, when the combined image was shifted left and right for the wiggle display, the upper level shifting overwrote the lower level information.  However having more holes to allow shifting destroyed too much information at the various level for meaningful weather displays.  Finally experiments were done with the images of only allowing significant information to be displayed and having the rest of the model image as blank.  Relative humidity fields showing the higher levels of RH worked well and looked similar to the wiggle satellite images.  Experiments with other fields will be shown at the conference.

References:

Mosher, Frederick R., 2012: “Revisiting 3D Stereo Satellite Image Displays”.  28^th Conference on Interactive Information Processing Systems (IIPS), American Meteorological Society, New Orleans, Jan. 2012.