It is possible to overcome the layered look of SIRDS by oversampling and averaging color values for pixels at the time of generation. Computationally this task is difficult due to echoes in the picture.
Another way to reduce this effect is to interpolate colors by using fractional parts of the stereo separation values. As a side effect the resulting images look a bit fuzzy.
Color SIRDS does not mean SIRDS with colored patterns, but the approach to put correct coloring into the 3D-picture behind it. It is very difficult to maintain colors for two differ-ent images in one without color distortion. The easiest way to bring color into the scene is by amplitude modulation: The color of a pixel is defined by the average of the colors of the im-ages for the two eyes. This method is relatively correct, but it leads to very distorted pictures. Better results can be obtained by using a more sophisticated averaging function which de-creases the saturation in that part of the picture when the difference of the two colors is large. Uniform coloring causes no change in saturation.
It is possible to create colored images, that loose their color when viewed, showing the 3D-picture in grayscale. This can be achieved by assigning complementary colors to both im-ages of a point.
There are two ways of using animation with stereograms:
First you can animate the random pattern of SIRDS (or the initial pattern of SIS). This eases the viewing, because the only static thing in the image is the contained 3D-picture.
The second possibility is to animate the 3D-image. The biggest problem is to maintain a continuous look of the background, because the colors of the pixels respond with strong varia-tions to small changes of the 3D-picture.
There are two ways of doing so:
It is possible to create rotation dependent RDS. In addition to the usual constraints in horizontal direction, constraints can be defined in an other direction (ex. vertically). Such a SIRDS shows two different pictures when rotated. Because of the resulting complex depend-encies probably not any two pictures can be combined in a single SIRDS.
It is also possible to create SIRDS showing different images viewed wall-eyed and cross-eyed [KURU 1994] (see the next chapter for an explanation of the viewing techniques). Using a chain-link method for the constraints (like our algorithm does) a wall-eyed SIRDS can be generated. In a second step the chains just have to be updated for the additional constraints of the cross-eyed image. The main disadvantage of this method is that is produces "foggy", diffi-cult to view images. An alternative method is to use pixels for cross and wall-eyed viewing alternating, which will at least reduce the horizontal resolution by 50%.
There is also no difficulty in putting several cross-eyed pictures into a stereogram. To ex-plain how to view them, the concept of "order of match of random buffers" has to be intro-duced. A match of random buffers of zero order equals to viewing the plane SIRDS image. The cross-eyed match of random buffers of nth order is shown in the following figure.
The "order of match" concept
In general, the greatest problem of multi-picture SIRDS is that they have lower quality and are more difficult to view than single picture SIRDS.
back to my homepage * back to the toc * continue with the algorithm c-code