<title>Quality measures for depth-from-focus approaches</title>

Abstract

Besides the ongoing rapid development of processing power of computers new standardised interfaces are emerging. Thus. future measuring software will be able to process information ofthe quality ofa measuring operation. This exceeds the current state-of-the-art. The originality of this research lies with the proposal of a novel method for computing a quality factor for each measuring point. The method is applicable for optical imaging sensors. However. similar regimes may be applied for other types of sensors. This paper presents a general methodology for computing a quality factor for focusing. Thereby different criteria such as unimodality. accuracy, reproducibility. definition range. general applicability and robustness are covered. The application of the proposed quality factor enables a more profound evaluation of a focusing process than the pure specification of the uncertainty of the focus position does. Within a closed quality loop it enables a higher level ofcontrol ofthe measuring process resulting in a significantly decreased measuring uncertainty and an increased robustness. Thereby the closed quality loop comprises the CAD process. inspection planning, measuring operations and the comparison between CAD data and measured geometry. The paper closes with some experimental results showing the soundness ofthe proposed method.

Enhanced low-light auto-focus system model in digital still and cell-phone cameras

“…The performance of the introduced LLAF system in this paper was compared with our previously developed LLAF system presented in [1], [2]. The SR improvement results can be seen in Tables 1 and 2 with the single focus value curve results denoted by FV and multiple focus value curve results denoted by MFV for TSR of 0.25.…”

“…For Camera 3 both methods achieved focusing down to 3 lux. By comparing the sharpness ratio results, for different target object contrast levels between the previous system in [1] and [2], and the introduced system, it is seen that the performance of the new LLAF system model with multiple focus value curves along with the preprocessing operator Tp [. ] exceeds the performance of the previous LLAF system by consistently lowering the lowest lux level at which AF can be achieved.…”

“…The quality of the sharpness measure is critical for the AF search algorithm to be able to locate the peak position quickly and accurately [1], [2]. The general consensus from the literature is that no one measure can adequately satisfy all the above criteria.…”

“…In low-light conditions, it is well known that the high-frequency content of the image becomes contaminated with noise, resulting in flat focus value curves. In [1], [2], we introduced a noise reduction filter prior to AF statistics calculation to elevate the peak of the focus value curve, thus enabling an AF search algorithm to locate the desired in-focus position.…”

“…This paper improves upon our previously introduced low-light AF system model which included a noise reduction preprocessing filter by considering an indexed family of focus value curves generated from a digital band-pass filterbank. Also, the Sharpness Ratio (SR) metric used to characterize low-light AF (LLAF) performance with a single curve in [1] and [2] is extended for use with multiple curves. Experimental results with three prototype cameras show that the improved AF system outperforms our previously achieved LLAF performance.…”