Stranica 26 [26]
OCR
CHARISM A I 7 ij The luminance matching between the calibrated visible-reflected images (such as that shown in Figure 1-14(b)) and the corresponding linearized UV- and IR-reflected images is carried out by comparison of the Spectralon diffuse reflectance standards included in these images. A matrix of the average X, Y and Z values for each of the standards is generated in both the visible-reflected and UV- and/or IR-reflected images and these are then matched by linear regression (See Chapter 3). Finally, the UVR and IRR images in XYZ space are made into monochrome images (“mono-ized”) with just the data from the X or Z channel, respectively. This is equivalent to taking the data from the B or R channels in RGB space, as described by Aldrovandi et al." However, as a tri-band image is still required, and the X Y and Z have different maximum values (X=95.047, Y=100.00, Z=108.883), the X or Z channels cannot simply be copied into the other two channels. Instead X or Z are scaled down to 0 - 1, and then scaled up again by the maximum of the channel it is being copying into. This process not only corrects for the camera-applied gamma and the spectral density of the radiation source but is a crucial step for creation of false-colour images, where the luminance of the images fed into the various channels must be well-matched. The integration of this approach into a workflow for the development of the post-processing software addressing the correction of reflected images is discussed in Chapter 3. The optimisation of experimental procedures, as well as the data acquisition requirements for post-processing, are discussed in Chapter 2. — Camera response (colour calibration for Visible-reflected images only) The colour signal captured by an imaging system, is the product of the incident light, the reflective properties of imaged surface and the spectral sensitivity of the recording device. The impact of spectral density on reflected images and the correction methods applied to address spectral inhomogeneity and camera applied gammas were discussed in the previous section. This section will discuss approaches towards the correction of the spectral sensitivity of the camera, commonly known as colour calibration. In DSLR cameras the RGB values recorded for an image are dependent on the spectral sensitivity of the camera, (as a result of its sensor and the characteristic response of the embedded R, G, B, IR cut-off filters), i.e. different sensors can yield different RGB outputs for the same scene. Figure 1-15 shows two visible-reflected images of a detail from a wall painting fragment as shot with two different cameras with different sensor types: a Canon 40D (CMOS sensor) and a Fuji S3 Pro (CCD-based sensor). Although subtle, a slight difference in colour can be observed between the two images, resulting from the differences between the spectral sensitivity of these sensors and other optical characteristics. Estimation of camera spectral sensitivity or response is necessary to colour calibrate visible-reflected images so that these may be reproducible and comparable. A standard technique for determining the spectral sensitivity of a camera establishes a relationship between incident narrow-band light with different wavelengths across the visible wavelength range and the camera output. The procedure used to achieve this is Summarised in Figure 1-16. A monochromator or narrow-band filters are used to generate a series of monochromatic wavelengths of light. Each of these is cast on a standard white Version No. 1.0 18 Date : 14/10/2013
