Segmenting Dead Sea Scroll Fragments for a Scientific Image Set

Bronson Brown-deVost,Berat Kurar-Barakat,Nachum Dershowitz

2024-06-22

Abstract:This paper presents a customized pipeline for segmenting manuscript fragments from images curated by the Israel Antiquities Authority (IAA). The images present challenges for standard segmentation methods due to the presence of the ruler, color, and plate number bars, as well as a black background that resembles the ink and varying backing substrates. The proposed pipeline, consisting of four steps, addresses these challenges by isolating and solving each difficulty using custom tailored methods. Further, the usage of a multi-step pipeline will surely be helpful from a conceptual standpoint for other image segmentation projects that encounter problems that have proven intractable when applying any of the more commonly used segmentation techniques. In addition, we create a dataset with bar detection and fragment segmentation ground truth and evaluate the pipeline steps qualitatively and quantitatively on it. This dataset is publicly available to support the development of the field. It aims to address the lack of standard sets of fragment images and evaluation metrics and enable researchers to evaluate their methods in a reliable and reproducible manner.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the automatic segmentation of manuscript fragments in Dead Sea Scrolls images. Specifically, the authors propose a customized processing flow to address the challenges faced by standard segmentation methods when dealing with these images. The following is a detailed problem description:

1. **Additional elements in the image**:
- Each image contains auxiliary information such as rulers, color bars, and plate - number labels. The positions of these elements vary in different images, which brings difficulties to segmentation.

2. **Similarity between the background and ink marks**:
- The image uses a black non - reflective background, which makes it difficult to distinguish between the background and the ink marks on the manuscript, increasing the difficulty of segmentation.

3. **Impact of modern support materials**:
- Many fragments are fixed on modern support materials, such as Japanese paper. This may lead to overlap or contact between the fragments and the support materials, further complicating the segmentation task.

To solve these problems, the authors propose a four - step customized pipeline. The specific steps are as follows:

1. **Isolate calibration bars and labels**: Use Faster R - CNN to detect and remove calibration bars and labels in the image.
2. **Align front and back images**: Align the front and back images through keypoint matching and the RANSAC method to avoid misidentifying ink marks as holes.
3. **Threshold processing**: Perform threshold processing on infrared images to generate preliminary fragment masks.
4. **Mask processing of background support materials**: Detect and remove the areas of support materials, and finally generate clean manuscript fragment masks.

In addition, the authors also created a publicly available annotated dataset for evaluating the segmentation effect and provided quantitative and qualitative evaluation results. This dataset includes color and infrared images of 139 fragments, as well as the corresponding segmentation ground truth, aiming to promote the research progress in this field.

Through this method, the authors hope not only to solve the current problem of Dead Sea Scrolls image segmentation, but also to provide references for other similar manuscript image segmentation tasks.