| 摘要 |
A computational image processing filter processes an image from a wedge-based imaging system so as to remove artifacts such as blurring and ghost images. By removing the artifacts computationally instead of optically, manufacturing costs and complexity are reduced over prior solutions. In one implementation, the computational image processing filter performs a two-dimensional transform to align a ghost image with a pixel grid defined by the wedge. The transformed image is then stretched using a nonlinear transform to make the ghost pitch versus position a constant. Next, an anti-ghost point spread filter is created and deconvolved. Finally, an inverse of the nonlinear mapping is applied. Artifacts introduced by other optical layers can be reduced by deconvolving the artifact from the image according to a point-spread function representing an effect of the optical layers on the image. |
