chapter 8, Flat Fielding

from: Photon Transfer

Author(s): James R. Janesick

Published: 14 August 2007

Chapter DOI: 10.1117/3.725073.ch8

Chapter Page Count: 15 pages

Previous Chapter | Next Chapter

Front Matter

1. Introduction

2. Photon Interaction

3. Photon Transfer Noise Sources

4. Photon Transfer Theory

5. Photon Transfer Curve

6. e⁻/DN Variance

7. Nonlinearity

8. Flat Fielding

9. Modulation Photon Transfer

10. Signal-to-noise Performance

11. Read Noise

12. Lux Transfer

A. PTC Data Reduction Example

B. PTC Simulation Program with Thermal Dark Current

C. PTC Simulation Program with FPN Removal through Flat Fielding

D. LTC Simulation Program with Thermal Dark Current

Back Matter

Preview

Chapter Contents

8.1 Theory
8.2 Photon Transfer Verification
8.3 Nonlinearity

Excerpt

8.1 Theory

Fixed pattern noise is removed from images by a technique called flat fielding, where a computer adjusts pixel sensitivities to be equal. Fixed pattern noise severely limits S/N performance for CCD and CMOS imagers, which will be discussed in Chapter 10. Fortunately, simple computer algorithms can remove FPN and achieve the shot noise limit, thereby significantly improving S/N performance. The image processing routine is based on the following linear equation:

where S_iCOR is the signal level of the corrected ith pixel without FPN (e⁻), S_iRAW is the signal level of the ith raw uncorrected pixel (e⁻), S_iFF is the signal level of the ith flat-field pixel used to remove FPN (e⁻), and μ_FF is the average flat-field signal level (e⁻). Equation (8.1) also can be applied directly to DN camera units as

For PT work, it should be mentioned that flat fielding can be substituted for the frame differencing technique illustrated in Fig. 5.3 to remove FPN. In doing so, the number of frames required to generate PTCs is reduced, and the flat-fielding process is verified for theoretical shot noise limited performance.

The flat-fielding technique demonstrated in Fig. 8.1 shows two raw sinusoidal video traces that are FPN limited (labeled as S_RAW). Also presented is a flat-field trace used to remove FPN (labeled as S_FF). The solid dark curves shown are the corrected traces after the raw traces are divided (pixel-by-pixel) by the flat-field level and the result multiplied by μ_FF according to Eq. (8.1). Note that S/N performance improves significantly after flat fielding. Figure 8.2 presents images for the lowest-contrast sinusoidal shown in Fig. 8.1 before and after flat fielding is performed. The improvement in image quality is obvious.

After flat fielding is performed, the corrected image contains “remnant shot noise” from the flat field itself.

DOI: 10.1117/3.725073.ch8

Your library does not subscribe to the eBooks portion of the SPIE Digital Library.