, , ,

Every image embodies a way of seeing. Even a photograph. For photographs are not, as is often assumed, a mechanical record. Every time we look at a photograph, we are aware, however slightly, of the photographer selecting that sight from the infinity of other possible sights.

—-By John Berger

Computer Vision (CV) and Digital Humanities (DH) complement each other in their individual progress. In the initial phase, computer vision played a significant role in the development of Digital Humanities. However, recently DH seems to make a promise for the progress of Computer vision. Let us see how CV helps in DH in archival image database and Script Analysis:

Archival Images

Arch-V (Archive Vision), a newly developed software providing automated, image based searching of the digital archives. It employs the state of the art method in image classification and recognition, from the field of Computer Vision. The methods in Computer Vision research is well derived for the contemporary real life applications. However, they have been adapted for the archival images.

The paper [1] outlined the method they have incorporated for digital archives and they claimed to have implemented their solution to the English Broadside Ballad Archive (EBBA) cataloging interface. The method involves the following steps:

  •  Bringing the colored image and black and white archival images to a common format for making them uniform.
  • Detecting notable feature points like corners, edges and among others.
  • Speeded Up Robust Feature (SURF) has been used as a feature to extract information around the detected feature points from raw digital archival images.
  • Visual Dictionary [5] of these features is created and given to supervised learning for making classification model.

Each image in the archive is passed through each of the steps and finally use the classification model to categorize the archival entity.

Script Analysis

Let us now see how computer vision helps in the handwriting analysis of scripts. It is quite different from the images we discussed above. It helps in the field of Paleography, study of ancient handwritten writing. The paper proposes a quantitative analysis framework [2].

The characters are represented as B-splines. The glyphic shape of a character is thinned and it is converted into splines. The feature extraction is not just done at this stage. More interestingly, we go into the trajectory reconstruction as we do not have the dynamic movements of pen for writing the particular script. Dynamic features can be observed by altering the trajectories which is very useful for defining the character. Furthermore, characters can be broken down into basic strokes for analyzing them i.e. Stroke segmentation.  This is performed on different significant points of the recovered trajectory of the character. Thus, we represent the characters as set of strokes which allows us to derive better features than just from pixelated data.

We saw the analysis of character for its production and realization, which give rises to following features: geometric, cognitive, and stroke Features. These computed features serves as a descriptor for the scripts. This can be used for comparing and analyzing scripts.

Role of DH in CV

The way Digital humanities help computer vision is quite different. The Study of digital humanities is of high significance because of the need for digitizing the history and extracting the information. This leads to better understanding and recognition of images in terms of context. How do we read a photograph? Though we can describe an image to have objects like house, toys, etc, the context of an image comes from the knowledge of history of the scene that the image captured. Every image will have some physical extension[3]. Eg: The picture of Wall of Jallianwala bagh is not just a portrayal of wall, however it depicts the struggle of Indians under British colonization. Whatever we see around in the world has a hidden historical significance.


Fig 1: Jallianwala Bagh, Amritsar, India [4]

The field of computer vision is moving towards the goal of recognition of objects and scenes and passing the Turing test. The readings of a photographs have various layers of recognition in which DH complements the higher layer of recognition. The CV field lack the knowledge of history heavily. The upcoming of Digital humanities will have a great impact on the overall understanding of the depiction of an image. I conclude by saying that Digital Humanities and Computer Vision have a lot of hope for progress joining their hands.


1. Stahmer, Carl. “Arch-V: A platform for image based search and retrieval of Digital Archives“, DH2014, Lausanne

2. Rajan, Vinodh. “Framework for quantitative analysis of Scripts“, DH2014, Lausanne.

3. Das Gupta, Vinayak. “Readings of a Photograph: Cognition and Access“, DH2014, Lausanne.

4. By Dr Graham Beards (Own work) [CC-BY-SA-3.0], via Wikimedia Commons

5. Reja Arandjelovic and Andrew Zisserman (2011). “Smooth Object Retrieval using a Bag of Boundaries“. Proceedings of the IEEE International Conference on Computer Vision (ICCV).