| Summary: | 博士 === 國立清華大學 === 電機工程學系 === 88 === In machine vision systems, color imaging plays a vital role in conveying and recording visual information from real-word objects. To accurately represent colors acquired from camera imaging, we propose techniques of spectral responsivity estimation and colorimetric modeling for the vision systems, in this thesis.
For determination of camera’s spectral responses, we introduce a multichannel filtering (MCF) system. The design objective of the optical system is to effectively select a limited amount of spectral (or broadband) filters to characterize the spectral features of color imaging processes, which are contaminated with noise, so that the spectral response functions can be estimated with satisfactory accuracy. In our approach, a theoretical study is first presented to pave the way for this work, and then we propose a filter selection algorithm based on the technique of orthogonal-triangular (QR) decomposition with column pivoting (QRCP), called QRCP-based method. This method involves QR computations and a column permutation process, which determines a permutation matrix to conduct the subset (or filter) selection. Experimental results reveal that the proposed technique is truly consistent with the theoretical study on filter selections. It is found that the MCF system with the filters selected from this method is much less sensitive to noise than those with other spectral filters from different selections. That is, the spectral responsivity estimation can achieve a satisfactory accuracy.
From the estimated spectral responses, we propose a colorimetric modeling technique to give a computational model associated with colorimetry, so that the representation of color acquired from camera imaging is accurate and meaningful. First of all, the colorimetric quality is evaluated to reveal the feasibility of this work. In the modeling process, we present a spectral matching method and an approach of determining a reference-white luminance. As a result, the acquired color and the true (or measured) color can be well coordinated, with the strength of a global illumination or display white, in a perceptually uniform color space, e.g., in CIE 1976 L*a*b* space (abbreviated as CIELAB). Then, lower-degree polynomial regression is employed to eliminate color errors due to the mismatch between spectral response functions. Experimental results indicate that the root-mean-square (RMS) value (i.e., color error) from the degree-3 polynomial regression is less than a just-noticeable difference (about 2.3) in CIELAB. It appears that the proposed technique can establish an accurate colorimetric model for machine vision systems.
|
|---|
