In order to prepare the image of a given fragment, unnecessary data must be removed. This is a prerequisite for any further graphic action, such as restoring of the fragment (chapter 3), filling the lacunae with letters (chapter 8), or placing the image in the digital canvas (chapter 12). The final goal is an image file that contains only the fragment on a transparent background. Graphic removal of the background constitutes a modern application of the Stegemann method. In her meticulous article on the principles of the Stegemann method, Steudel recommends drawing the boundaries of each fragment by hand, disregarding any unnecessary data which does not belong to the fragment and originates from the background.1 Following Steudel’s recommendation, scholars should use digital tools that yield accurate and precise results with regard to minute details. This chapter discusses the digital process of background removal. Following a brief discussion of the methodological aspects, a detailed technical manual of the process is provided.
Producing an image file, clear of unnecessary data from the old PAM images may present challenges to the scholar. For example, the old PAM images often present many fragments together on a single plate, plus a variety of additions, such as the scale, notes, and adhesive paper. Another problem is that the skin and the shadow it casts on the plate are quite similar, making it difficult to distinguish even for the human eye. The new LLDSSDL images (figure 5) capture the fragments with modern additions such the “checkers” plate, the scale ruler, the plate number tag, additional Japanese paper or adhesive tape that support the fragment, all on pitch black background.
Figure 5 shows a fragment that is imaged on a black background with Japanese paper supporting the fragment in the lacunae.
Before initiating any digital reconstruction, the modern trappings of the image should first be removed. The simple way to do this is to use one of the many online background removal tools.2 These tools can assist in removing the majority of the background, but at present they cannot be fully trusted to only or entirely erase the unnecessary parts of an image. The main problem that scholars may encounter is the similarity in the ink and background colors. Where the letters abut on the background, automated tools may mistakenly identify them as part of the background and remove them accordingly. In addition, some tools may reduce the image resolution and change its scale. The SQE platform applies to each fragment a mask calculated by an advanced algorithm, which in the great majority of pieces captures the borders of the fragment quite accurately.3 Nevertheless, a careful human eye is still required to validate the automated procedure and correct it when necessary. The platform also provides a handy way for correcting the mask of unproperly separated fragments, usually in the case of pitch black pieces of skin.
1 Manual Removal of the Background
The first step of the process is a quick demarcation of the desired fragment with any of the selection tools. One should then cut out the fragment and paste it as a new image, disposing of all unnecessary data around it (figure 6). In some programs, such as GIMP, deleting the surroundings of the image will result in an undesired white background, rather than a transparent one, hence the need to cut the fragment and paste it in a new file.
Whatever is left of the background must now be removed. Since the color of the background resembles the ink, special attention is needed in order to not erase any signs of ink during the process, particularly for letters whose ink abuts the background. The removal can be done in several manners. Using the Fuzzy Select/Magic Wand tools requires some experimentation in order to find the proper threshold. Even so, operating these tools may conclude with minute black dots surrounding the fragment, which harms the quality of the result. We therefore advise the use of the Scissors Select (GIMP) or Scissors (PS) tool, which will allow scholars to manually cut out the remains of the background and the Japanese paper (figure 7).
The final step is the removal of the fine details of Japanese paper that stand within the fragment. This step is also the most delicate one. The unique texture of the paper makes it difficult to remove with smart selection tools and requires the use of a free selection tool, carried out manually (figure 8).
The procedure described so far applies to the LLDSSDL images, but its application to the older PAM images is more complex. Unfortunatly here there is no easy solution. The low resolution of the old images and the frequent shading surrounding the fragments require careful and slow manual work with a free selection tool, until the desirable results are reached.