Because of the limited resolution of computer screens and other output devices only a small number of different levels of depth can be portrayed. Increasing the resolution of output devices does not solve this problem. If the dots get too small, the eye has problems to recog-nize the corresponding parts of the image and to get the depth impression.
Assuming an eye separation E=2.5 inch, m=1/3, and a screen resolution of 75 DPI we get a stereo separation of 93 pixels for the far plane, and 75 for the near plane (resulting from s=E(1-mz)/(2-mz) evaluated for z=1 and z=0). That means that the scene will be displayed using just 93-75+1=19 different depth levels. It will probably look like if it was built of several layers of lego-stones. In addition the distance between two layers is not constant, because of the non-linearity of the stereo separation function.
All SIRDS and SIS-pictures suffer from fuzzy edges. The reason is that pixels at the left and right edge do not have corresponding pixels for the other eye. The eye only receives "mono" for that areas of the scene.
Echoes are points that appear in the 3D-scene because of incorrect fusion of pixels. Peri-odicities within the image can be interpreted as non existing objects. In fact, every two equally colored pixels separated by less than the far-separation can be interpreted as the images of a point of the scene. The stronger the constraints within a image, the greater the chance to pro-duce echoes. Hidden surface removal (hsr) in our program reduces constraints, lowering the amount of echoes. Echo reduction for SIS and SIRDS generated using a repeating pattern is obtained by starting generation in the middle of the image. Artifacts are echoes originating in pixels given the same color by chance. They can easily be eliminated by rapidly changing the random pattern while viewing (shimmering, pattern animation) or by the use of gray levels in-stead of just black and white.
Echoes in an image generated with our program without use of hidden surface removal
SIRDS of the same object generated with hidden surfaces removed.
The pictures above show echoes within SIRDS generated with and without hsr. The scene is a hollow rounded rectangle placed in the near plane. A white pixel indicates, that two equally colored pixels separated by the stereo separation of near were found in the SIRDS at this posi-tion (x and x+near). This is of course true for the image of the rectangle. The noise around it appears because of random matches of images of the background. The left figure, generated without hsr also shows a large and several small echoes (uniform white areas).