Cambridge MSI Data

In collaboration with the Cambridge University, the Gower manuscript R.3.2 in the Trinity College was multi-spectrally imaged with a 8-channel multi-spectral camera. The whole procedure of producing the ready-to-use data is composed of two steps described as follows.

Step 1: Data Acquisition

The focus of this step is on obtaining the raw image data of the pages of the manuscript R.3.2, as well as some extra data, which shall be used for calibration purpose in the following step of data processing. Following the process below, we acquired the data.

1) Set the manuscript in correct position on book cradle and adjusted camera position and focus, then removed book.

2) Adjusted light position to obtain homogeneous lighting across a white card laid across the entire field of view. Light levels at the object are approximately 600‐650 lux

3) Adjusted exposure for each filter by checking signal levels on black and white reference standard (white fluorilon, black spectralon). The maximum exposure allowed by the imaging system is 2000 ms, which does not allow to satisfactorily maximize signal levels for a couple of the filters used, especially filter 7 (CWL 375 nm).

4) Captured image set for white card and a “dark” image set (captured with lenscap on).

5) Acquired reflectance spectra as well as image set of black and white reference standards placed within field of view of camera.

6) Captured image sets for white reference standard placed in 58 different positions in order to cover entire field of view of camera.

7) Captured image sets (bit-depth: 8) for each page in the book.

The below shows the list of filters and the exposure used for each filter.

FILTER    CWL    BP    Start    Stop    Exposure (ms)
7             375      50     350       400      2000
8             425      50     400       450      2000
1             475      50     450       500      1100
2             525      50     500       550      850
3             570      50     545       595      600
4             625      50     600       650      750
5             680      50     655       705      800
6             800      LP     500                        
Step 2: Data Processing
Given the data from Step 1, this step is to post-process it, producing the ready-to-use data. The processing includes two aspects, which we refer to as “internal calibration” and “external calibration”. The former is to remove imperfections due to the camera itself; by contrast, the latter is a white balance calibration. 
The internal calibration is carried out, for each channel, according to the equation [1]:
                                                                             IC = [(IR - IB) * M] / (IF - IB)
where IC is the calibrated image; IR is the non-calibrated object exposure; IB is the bias or dark frame; M is the average pixel value of the corrected flat field frame; and IF is the flat field frame.
After calibrating all the raw data, we can then perform the external calibration by employing
                                                                             IC’ = IC * cos(θ) * r / IS * M
where IC’ is the final calibrated image; θ is the angle between the light direction and the normal to a page of the book; r indicates the spectralon response; IS is the calibrated spectralon image and M is the maximum pixel value (M = 255 for an 8-bit image).
Given a set of 8 calibrated images for a page, a single visualization file, represented as .exr, can be created, and it can be visualized using the open-source software CHER-Ob developed by the Yale Computer Graphics Group