How to Measure a Camera's Response Curve from Scratch?

Martin Bichsel and Krystyna W. Ohnesorge
University of Zurich
Department of Computer Science
Multi-Media Laboratory
Winterthurerstr. 190
8057 Zurich-Irchel

Abstract

This paper describes how the response curve of a video camera can be determined with the aid of a sheet of paper and two desk lamps only. The calibration process is based on the superposition law of incoherent light and does not make use of any calibrated measurement instrument.

1 Introduction

Digitizing and processing of video images has steadily gained importance in recent years. Various applications profit from this technique, in particular industrial inspection, optical measuring, desktop publishing, and remote sensing. For applications which are sensitive to image quality an exact characterization of the properties of the data acquisition equipment is essential. Knowing the equipment's response curve is especially important because many applications either rely on a linear response curve or on a specific power law ( -correction). Relying on the data sheets of the measurement equipment is only partially useful because

1. The data is often missing (e.g. for the digitizer).

2. The data sheets may be incomplete, inaccurate, or even erroneous.

3. The measurement equipment as a whole (i.e. camera plus digitizer) may react in a way that is not expected from the individual parts, e.g. due to undesired feedback.

Calibrating the system with the aid of a calibrated measurement instrument is the standard procedure. In some situations, however, the user would like to be completely independent of any other measurement instrument because

1. He does not trust the calibration instrument, which may be imprecise or even broken.

2. Calibration instruments are expensive.

This paper presents a measurement method which provides precise calibration results and only relies on equipment that is readily available in any laboratory or factory: the system we want to characterize (i.e. a camera, a frame store, and some computer), two light sources (desk lamps or spot lights), and a sheet of uniformly white or gray paper.

The proposed method bases on the additivity of incoherent light (Section 2) and on the fact that various sets of light sources easily can be adjusted such that an equal system respons results (Section 3), even if the exact response curve is not yet known.

?This work was supported by the consortium VISAGE, KWF grant No. 2440.1 (first author), and Swiss National Science Foundation grant #21-29915.90 (second author).