Effects of Camera Position and Media Type on Lifelogging Images (MUM Best Paper Award)

At this year’s International Conference on Mobile and Ubiquitous Multimedia (MUM) we presented results from a RECALL study on lifelog camera  positioning. MUM is a leading annual international conference, which provides a forum for presenting the latest research results on mobile and ubiquitous multimedia. The paper – spearheaded by Katrin Wolf – has been awarded with the best paper award.
In the paper we report on an evaluation of existing visual lifelogging capture approaches. Specifically, we conducted a user study with two main capture dimensions:

  1. Comparing the body position where a lifelogging camera is worn: head versus chest.
  2. Comparing the media captures: video versus stills.

Therefore, we equipped 30 participants (Fig. 1) with cameras on their foreheads as well as chests, and later asked them about their perception of the images collected. Additionally, we applied a set of standard image processing algorithms to classify images, such as sharpness filters or face and hand detection.

Participant equipped with cameras on head and chest.
Figure 1. Participant equipped with cameras on head and chest.

Our findings indicate that

  1. Chest-worn devices are more stable and contain less motion blur through which feature detection by image processing algorithms works better than from head-worn cameras.
  2. Head-worn video cameras, however, seem to be the better choice for lifelogging as they capture more important autobiographical cues than chest-worn devices, e.g., faces that have been shown to be most relevant for recall.

Our work contributes to the domain of lifelogging image capture by showing that video provides more beneficial information than stills, while head-worn cameras capture more subjectively valuable content, e.g. faces. Hence, a head-worn video camera seems to work better as a lifelogging camera device. The paper provides a comprehensive overview of lifelog camera images and gives further pointers as to when to select certain camera devices, positions, and media types.

For more detail and future reference, please see:

Katrin Wolf, Yomna Abdelrahman, David Schmid, Tilman Dingler, and Albrecht Schmidt. 2015. Effects of camera position and media type on lifelogging images. In Proceedings of the 14th International Conference on Mobile and Ubiquitous Multimedia (MUM ’15). ACM, New York, NY, USA, 234-244. DOI=http://dx.doi.org/10.1145/2836041.2836065