Customizing imagery to improve visual acuity

Ignacio Garcia Dorado, Purdue University

Abstract

This thesis describes a new technique that modifies an image (i.e., the pixel values) such that an individual who has a refractive vision problem perceives it as sharp and, and so has less dependence on corrective eyewear. Since this approach varies only the pixel values, it can be presented on any conventional digital display (e.g., computer screen, smart-phone) or printed media, avoiding the use of any custom hardware. The method is based on a constrained total-variation method deconvolution that reverses the effect of imperfect vision. In the deconvolution process, a novel relative total variation term is introduced that controls the ringing (i.e., undesired ``halo'' effect generated by the deconvolution). The end result is the ability to generate sharper appearing images for individuals with refractive vision problems including myopia, hyperopia, presbyopia, and astigmatism. This approach has been validated in simulations, in camera-screen experiments, and in a user study.

Degree

M.S.

Advisors

Aliaga, Purdue University.

Subject Area

Computer science

COinS