Proc. SPIE. 7496, MIPPR 2009: Pattern Recognition and Computer Vision
KEYWORDS: Mathematical modeling, Eye, Visual process modeling, Modulation, Spatial frequencies, Retina, Human vision and color perception, Modulation transfer functions, Contrast sensitivity, Eye models
This paper is a study, based on the limitation of human vision characteristic, of image recognition through the take
account of correction factor. Those aspects that have been explored focus on human eye modelings, including human
vision recognition characteristics and various mathematical modeling verify. By using Modulation Transfer Function
(MTF) curve evaluation recognition capability on the studied models, an optimum recognition model most compatible
to human eye physiology is summed up.
This paper proposes a new method for optimization optics with a diffractive optical element (DOE) via a Hybrid
Taguchi Genetic Algorithm. A Diffractive Optical Element, based the theory of wave phase difference, takes advantage
of the negative Abbe number which might significantly eliminate the axial chromatic aberrations of optics. Following
the advanced technology applied to the micro lens and etching process, precisely-made micro DOEs can now be
manufactured in large numbers. However, traditional least damping square has its limitations for the optimization of
axial and chromatic aberrations with DOE. In this research, we adopted the genetic algorithm (GA) and incorporated the
steady Taguchi method into GA. Combining the two methods produced a new hybrid Taguchi-genetic algorithm
(HTGA). Suitable glass combinations and DOE positions were selected to minimize both axial and lateral chromatic
aberration in the optical system. This new method carries out the task of eliminating both axial and lateral chromatic
aberration, unlike DOE optimization by LDS, which works for axial aberration only and with less efficiency. Experiments show that the surface position of the DOE could be determined first; in addition, regardless of whether chromatic aberration was axial or longitudinal, issues concerning the optical lens's chromatic aberration could be significantly reduced, compared to results from the traditional least damping square (LDS) method.
Chromatic Aberration plays a part in modern optical systems, especially in digitalized and smart optical systems.
Much effort has been devoted to eliminating specific chromatic aberration in order to match the demand for advanced
digitalized optical products. Basically, the elimination of axial chromatic and lateral color aberration of an optical lens
and system depends on the selection of optical glass. According to reports from glass companies all over the world, the
number of various newly developed optical glasses in the market exceeds three hundred. However, due to the complexity
of a practical optical system, optical designers have so far had difficulty in finding the right solution to eliminate small
axial and lateral chromatic aberration except by the Damped Least Squares (DLS) method, which is limited in so far as
the DLS method has not yet managed to find a better optical system configuration.
In the present research, genetic algorithms are used to replace traditional DLS so as to eliminate axial and lateral
chromatic, by combining the theories of geometric optics in Tessar type lenses and a technique involving Binary/Real
Encoding, Multiple Dynamic Crossover and Random Gene Mutation to find a much better configuration for optical glasses. By implementing the algorithms outlined in this paper, satisfactory results can be achieved in eliminating axial and lateral color aberration.
Advances in digital image optics have increased the significance of lateral color aberration because it is easily seen in the projected area. The choice of optical glass plays a role in the elimination of lateral color aberration. Current optical software still has difficulty in finding the optimal combination of optical glasses for twelve or more elements in a projection lens, the choice being among at least 300 optical glasses that have been developed. Even the modern damped least squares, a ray-tracing-based method, is limited, owing to its inability to identify an enhanced optical system configuration. As an alternative, this research proposes a new optimization process by using algorithms involving the theory of geometric optics in a projector lens, real encoding, multiple dynamic crossover, and random gene mutation techniques. Results and conclusions show that attempts to achieve negligible axial and lateral color aberration are successful.