Optimizing performance of autostereoscopic lenticular displays can be achieved by altering specific interdependent design parameters, e.g., width and number of views, screen disparity and lenticular slant, resulting in different crosstalk distributions and amounts of banding and consequently, different percepts. To allow the evaluation of an autostereoscopic lenticular display, before a costly physical sample is produced, an emulator was build. This emulator consisted of a goggle-based striped polarized display, a camera-based head tracker and software for generating L/R stereo pairs in real-time as a function of head-location. This paper addresses the development of the emulator, its validation with respect to an existing physical prototype, and the perceptual evaluation of three emulated fundamental design extremes: 1) a 9-view low-cross-talk system; 2) a 9-view intermediate crosstalk system; and 3) a 17-view high crosstalk system.Description: To demonstrate the performance of the emulator, a video was captured while moving a camera through the 15 views of both the emulator and the Quad Full HD autostereoscopic lenticular display. The camera was translated parallel to the display plane with the optical axis of the lens orthogonal to the display plane. The resulting video clips clearly show that position-dependent cross-talk introduced by differences in the sub-pixel visibility and visible as depth-dependent blur was adequately modeled as well as the perceptual effect of cone transitions.
© 2012 IEEE
Marc Lambooij, Karel Hinnen, and Chris Varekamp, "Emulating Autostereoscopic Lenticular Designs," J. Display Technol. 8, 283-290 (2012)