EE 267 - Project proposal

Three dimensional (3D) displays are becoming popular consumer technologies. The 3D effect mainly comes from the projection of two respective images on right and left eyes. For example, some movie projectors superpose two different images on the same screen with orthogonal polarlization states so that viewers wearing polarlization glasses see different images with different eyes. Head-mounted displays exploit lenses with a near-eye display to directly display separate images to each eye. In the expense of their high-quality 3D effect, however, they require specialized hardware. Among such 3D technologies, anaglyph is the most inexpensive way for the projection of two different images. This method uses a conventional display or a printer and plastic red-cyan glasses for the separate projection. The display shows an image only with a red channel for a left eye and an image with green and blue channels for a right eye. The red-cyan glasses wore by the viewer block out the respective channels for each eye to decompose the superposed image. Although this simple setup provides a 3D effect, it lacks color preservation for both eyes due to the color filters. This lack is the main disadvantage of anaglyphs among others including color distortion, retinal rivalry and ghosting effect. To suppress these disadvantages, several algorithms have been developed for generating anaglyph images. In 2001, Dubois developed a least square method to minimize the color discrepancy between original stereoscopic images and generated anaglyph images in the CIE XYZ color space [1]. This XYZ anaglyph method significantly improved the color distortion and has been considered to be the state-of-the art method until recently. Following this work, McAllister et. al. performed the minimization of Euclidian distance in the CIE L*a*b* space, which produced better color perception due to the perceptual linearity in CIE L*a*b* space [2]. However, this L*a*b* anaglyph method is computationally expensive because it solves an iterative optimization problem for each pixel. To further improve the color perception, Li et al. proposed a method to minimize the color discrepancy in HSV color space [3]. In addition to its superior performance, their method significantly reduced the computational cost by finding an approximated closed-form solution for the minimization. Similarly, this type of color retaining algorithm has been developed for a projector in which color primaries are customized [4].