Paper Reading #3: MUSTARD: A Multi User See Through AR Display

Intro:

• MUSTARD: A Multi User See Through AR Display
• Karnik, Abhijit,Walterio Mayol-Cuevas, and Sriram Subramanian. (2012).  MUSTARD: A Multi User See Through AR Display. Proceedings of the 2012 ACM annual conference on Human Factors in Computing Systems (CHI 2012), 2541-2550.
• Author Biographies:
• Abhijit  Karnik is a PhD student in the Interaction and Graphics group at Bristol. He studies and researchers multi-user and multi-view displays. His research is funded by Microsoft and he began interests in CHI in 2009. 
• Walterio Mayol-Cuevas  is a deputy at the Robotics Lab at Bristol. His research interests include personal robotics and real time vision groups. 
• Sriram Subramanian is a professor of Human-Computer Interaction at Bristol. He worked as a senior scientist at Philips Research in the Netherlands. Previously, he was assistant researcher at University of Saskatchewan

Summary:
The essence of this research is to deliver viewer information behind a cabinet. The information is constructed through multiple user input. Two challenges arise within the idea of an augmented reality such as delivering realistic visibility of the physical object behind semi-transparent layer and simultaneously relying view dependent information to the user.

MUSTARD allows viewers to inspect objects behind a glass panel while displaying the physical object through a projection. The system consists of two liquid crystal elements, a dynamic hole-mask and a data-panel. The dynamic hole-mask was chosen over a static mask for a plethora of reasons, but the figure below demonstrates the advantages of the dynamic mask through the polarization of physical objects laying next to the data layer. An implementation issue arises when the LC generates a twisting action which effects the polarized light passing through the front polarizing end. A picture summarizing the technique in which MUSTARD functions is given below.

The key concept of MUSTARD is to assess different views to different users, but using the same data panel. This creates an immensely usable display purpose as output. The algorithm works by applying mask-rendering, which uses the hole-mask, an image composed of black and white dots. However, the white dot locations must change position so that the white dots cover the entire display area in 10 frames. The next step involves generating composite view image, which essentially constructs a single image from multiple sources by calculating their point of view and the actual associated data. The last step involves conflict management which is the part of MUSTARD that involves determining overlap from the multiple view points and sorting through the mix. A symbolic picture of the implementation is viewed below.

In conclusion, the MUSTARD offers some intriguing possibilities for the future of physical interaction through a barrier, such as at a museum. The evaluation of the research conducted in this report can be read below in a subsequent section.

Related work not referenced in the paper:
1) "Towards Massively Multi-User Augmented Reality on Handheld Devices" by Wagner, Pintaric, Ledermann, and Schmalsteig
2) "Augmented reality meeting table: a novel multi-user interface for architectural design" by Penn, Mottram,  Schieck, Wittkämper, Störring, Romell, Strothmann, and Aish
3) "MARE : Multiuser Augmented Reality Environment on table setup" by Grasset and Gascuel

4) "Multiple Head Mounted Displays in Virtual and Augmented Reality Applications" by Kaufmann and Csisinko
5) "Collaborative Augmented Reality: Multi-user Interaction in Urban Simulation" by Ismail and Sunar
6) "Augmented reality interactive exhibits in Cartographic Heritage:  An implemented case-study open to the general public" by Grammenos, Zabulis, Michel, and Argyros
7) "Replicating augmented reality objects for multi-user interaction" by Lenting
8) "Visuo-Haptic Collaborative Augmented Reality Ping-Pong" by Knoerlein, Szekely, and Harders
9) "Distributed Augmented Reality for Collaborative Design Applications" by Ahlers, Kramer, Breen, Chevalier, Crampton, Rose, Tuceryan, Whitaker, and Greer
10) "Interaction Management for Ubiquitous Augmented Reality User Interfaces" by Hilliges

The research posed in these papers definitely exceeded the novel requirements. The vast majority of these works relied heavily on developing an augmented reality. This is an interesting concept that will pose some intriguing questions as organizations, such as the military, attempt to turn this research into industrialization. Furthermore, the related works was relevant to each other, and in fact, posed some of the same questions and problems, but in separate domains. Thus, this is a new a growing field which was major implications for the advancement of technology. The main difference that this paper focuses on is the augmented reality with multiple users in association. Furthermore, the MUSTARD research's reality adheres to almost a tangible nature potential of the outputted image.

Evaluation:
The researchers evaluated the difference in output of the image by comparing the test-source with the expected output in the absence of the reference-source. The high conflict scenario occurs when all users are viewing similar data, but at very different angles of examination. This problem results in distorted output that have significant overlap between multiple images, although it was recognizable in simplistic cases.

For other cases, the researchers use peak signal to noise ratios. This methodology analyzes the extraneous signals in the output, which may distort the actual images that a user of this technology would see. The PSNR ranged from 125 to 138 dB, in the high conflict case. The researchers graph the PSNR for each case in differentiating scenarios, for a total of  120 tests.

Overall, the researchers did an effective job evaluating results. The bottom line to compare images was very subjective and biased from the researchers point of view. However, to break it down into quantifiable terms, the researchers realized that PSNR was the most efficient method in determining the error. Thus, they were able to quantify this measure, and then successfully compare and contrast a multitude of trials against one another.

Discussion:
The concept behind this research was extremely novel and relevant to advancing technology. This evolution of this research idea could lead to an entirely new technology with an abundant number of potential uses. The user experience, from trial runs, tended to be positive. In my opinion, the research was very insightful. The researchers could definitely expand upon this. However, to improve their research paper, more images would have coerced with the definitions in order to fully explain the background of MUSTARD. All in all, the research was interesting enough that I will be checking back on the authors for follow-up work.