Camera Calibration Technical Report:
Introduction

Introduction

Techniques of microdensitometry for measuring the photographic or optical density of staining of microscopic sections are well established and accurate. The most common method to establish the integrated optical density of a sample of tissue, for example cell nuclei, is to sum the density measurements made using a pin-hole and photomultiplier tube scanned across the specimen mechanically. Whilst this method will usually result in higher accuracy it has the serious limitation of scanning relatively slowly. In contrast, video cameras can gather data very quickly and are cheap and convenient to use, particularly since the development of CCD array camera technology.

A number of authors [1,2,3,4] have reported the use of video cameras for densitometric measurements in a wide variety of applications. All these papers describe calibration techniques which either require a full set of neutral density filters or assume that the camera response is perfectly linear with respect to the incident light intensity. Furthermore these methods do not take account of shading (variation of illumination intensity over the imaged field) or the dark field response of the camera except by restricting measurements to a small square in the center of the field, typically 1/4 to 1/16 th of the full image area or by a simple subtraction of the bright field value.

In particular the linearity of the camera response is widely regarded as essential [4] for densitometry and indeed most measurement systems assume that the transmitted intensity is directly proportional to the camera output voltage. Although CCD array sensors may be accurately linear in themselves, the final measured value depends on the amplification within the camera, which often incorporates a $\gamma$ adjustment, and on an analogue to digital conversion. In any case, as most digitising systems will digitise the analogue signal to only eight-bit grey-level resolution (ie 256 discrete levels) it is not always desirable to adjust the camera to operate linearly. There may be a significant advantage in increasing the resolution over parts of the intensity range at the cost of reducing it over others so that maximum useful information for a particular application may be encoded into a restricted digital range.

In this paper we report a calibration and image capture technique that applies to cameras with arbitrary response functions, properly taking account of shading and the dark field response of the camera. The method is quick, efficient and requires only a single neutral density filter which is to be regarded as a reference standard. In the next section we detail the theory of the method and in the following section show how the theory can be implemented as a simple and efficient calibration technique. The latter section also considers the treatment of uneven illumination or shading.