Centers/Research Area Affiliations
Dr. Shrinivas Pundlik’s main research interests include image processing, computer vision, vision science, vision rehabilitation, and ocular biometrics. After completing his doctoral studies in the electrical engineering, Dr. Pundlik developed novel image processing algorithms for biometric identification using iris and face images. Since 2010, Dr. Pundlik has been engaged in various research projects that involved the application of computer vision and machine learning to develop new technologies and solutions for vision screening and vision rehabilitation. He also developed a wearable collision-warning device to help people with severe vision impairments avoid colliding while walking. He has contributed to the development of numerous mobile apps, either intended as low-vision aids or vision measurement for healthcare professionals. His research has resulted in more than 20 peer-reviewed publications in international journals and conferences, one patent, and a patent application under review.
Download his CV [PDF] for more information.
M.S. and Ph.D., Electrical Engineering, Clemson University (2005 and 2009)
Research Associate, School of Computing, Clemson University (2009-2010)
Postdoctoral Fellowship, Schepens Eye Research Institute of Mass. Eye and Ear (2010-2016)
2018: National Eye Institute Travel Grant, Association for Research in Vision and Ophthalmology
2015: Outstanding Reviewer Recognition, IEEE Transactions on Instrumentation and Measurement
2007: Harris Outstanding Teaching Assistant Award, Electrical and Computer Engineering Department, Clemson University
While conducting his Ph.D. research, Dr. Pundlik focused on developing computer vision algorithms for segmentation of images and videos (or clustering or grouping based on a particular property) for automated video processing applications. Dr. Pundlik proposed a fast algorithm for segmenting videos based on image motion that was able to automatically determine the number of distinctly moving regions in the image, dynamically add or remove a region as it appeared or disappeared in a video, and split a cluster into multiple parts if needed.
After earning his Ph.D., Dr. Pundlik developed algorithms for biometric identification using iris and face images. His research focused on processing images of faces, irises, and the region surrounding eyes (periocular region) to extract discriminative features to improve recognition accuracy. His work showed that the region immediately around the eyes (periocular region) is one of the most discriminative regions of facial images.
Dr. Pundlik joined Schepens Eye Research Institute of Mass. Eye and Ear in 2010, where he has been conducting research within the areas of vision science and ophthalmology, specifically in vision rehabilitation, mobility enhancement of visually impaired people, and vision-related diagnostics. In relation to the area of mobility enhancement, Dr. Pundlik helped develop a video-camera-based, wearable collision warning device for individuals with blindness and visual impairments. The warnings are provided via audio beeps or vibratory wristbands. The computer vision algorithms running on the custom, low-power device hardware analyze patterns of optical flow within the video in real time. A patent related to the collision prediction algorithm was granted in 2017.
Dr. Pundlik has also been involved in the development of vision rehabilitation smartphone applications. One of the applications is related to making smartphones more accessible to low-vision users. Many low-vision users of smartphones rely on screen magnification to view the contents on the device. However, accessing magnified screen content via manual scrolling is difficult and tedious, often resulting in inefficient usage of the app. A novel solution to tackle this issue was to wirelessly project the magnified smartphone screen onto the Google Glass head-mounted display so the user can control the part of the magnified screen visible with his or her head movements. This is equivalent to looking at a large virtual image through a head motion controlled viewing port, in this case, the Google Glass display. Remote interaction with the smartphone screen was possible with tap gestures on the Google Glass. The system was evaluated for routine tasks performed on a smartphone device. It was determined that using Google Glass head mounted display saves a significant amount of time compared to manual scrolling, when interacting with magnified screens that have familiar screen content.
Dr. Pundlik has also contributed to the development and evaluation of the SuperVision+ Magnifier. This mobile app was developed to help people with low-vision view details in the world around them, such as reading text and searching for things at a distance. Just like any smartphone screen, a simple pinch gesture enables the users to zoom in to magnify the live scene captured by the mobile camera, with an option of freezing the picture. Users are also able to stabilize the magnified live video by simply touching the screen. Studies showed that the app helped users, especially those with low-vision, read dynamic distant text.
Since 2013, Dr. Pundlik has been involved in the development of mobile apps intended for use in the clinical measurement of various vision-related parameters that aid in the diagnosis of vision disorders. He has been involved in the development of a mobile application for objective measurement of strabismus to help clinicians determine suitable treatment options. The smartphone app offers an accurate, rapid, and relatively easy way to obtain a measurement of eye alignment by taking a picture or a short video with the smartphone camera. There is no need for explicit calibration and or any other external attachment to the smartphone, which makes it easy to use and relatively inexpensive. Smartphone apps can also be designed to work on generic smartphone virtual reality headsets, and it offers many interesting possibilities in clinical testing and measurement of vision functions. Dr. Pundlik has helped develop a smartphone VR app for performing the sensorimotor tests in patients with brain injuries. Mobile healthcare apps offer a great potential to make vision screening accessible to a wider population. Dr. Pundlik’s recent work is a step in fulfilling that potential.
- Mobility and vision rehabilitation
- Computer vision
- Vision science
- Mobile healthcare apps
Wearable Collision-Warning Device
People with visual impairments, especially those with peripheral vision loss or blindness, often experience mobility related challenges and have a high risk of collisions or falls while walking. Dr. Pundlik is developing a wearable video camera-based collision warning device that detects impending collisions by processing videos captured from a body-mounted camera in real time, and provides audio or vibro-tactile warnings. The device computes collision risk based on the time-to-collision and the collision trajectory of objects in the surrounding environment captured by the body-mounted camera. The device has been shown to be effective in lowering the number of collisions in an indoor obstacle course in people with severe visual impairments and peripheral vision loss. Evaluation of the device in a long-term home use study is ongoing.
Mobile App for Measuring Strabismus
Strabismus, or eye misalignment, affects about 3-8 percent of Americans. Children younger than 10 years of age and older adults with head trauma due to stroke or injury are primarily at risk of developing strabismus. In children, untreated eye misalignment can lead to amblyopia (severe loss of vision in the misaligned eye). In adults, it can lead to double vision, spatial confusion, and diminished depth perception, if left untreated. Early screening in children can substantially reduce the risk of developing vision defects. In adults, an objective measurement can facilitate appropriate intervention to alleviate double vision. However, there is an unmet need for a rapid, convenient, accessible, and inexpensive device that can provide objective assessment outcomes to clinicians. Dr. Pundlik has developed a mobile app to measure eye misalignment that allows an examiner to capture image(s) of the patient from a smartphone and provides quantifiable measurements of strabismus. Further development, addition of various operating features, and clinical testing of the app is ongoing.
Smartphone Virtual Reality Apps for Vision Rehabilitation
Brain injury due to trauma or stroke sometimes affects the nerves that control movements of the eyes that results in misalignment of eyes only at specific gaze directions such as when looking below or above. Measurement of eye deviation at different gaze points can be important to determine the underlying cause and effective intervention to alleviate the problem. The smartphone virtual reality headsets provide an inexpensive platform for performing the sensorimotor tests of the eyes. Dr. Pundlik has developed a virtual reality app in which the patient performs a vision test of aligning two stimuli presented dichoptically while wearing the virtual reality headsets. This makes administering the test efficient and more repeatable. Preliminary research found that the app can accurately measure eye deviation at different gaze points in simulated scenarios. Dr. Pundlik is currently performing an extended evaluation of this app in outpatient and rehabilitation clinics.
Understanding of Saccadic Eye Movements in Naturalistic Viewing Tasks
Saccadic eye movements, or saccades, are fast ballistic movements of the eyes between fixations (steady gaze). Saccades are characterized by latency, peak-speed, amplitude, and duration. Under controlled viewing conditions, the saccades parameters generally tend to follow a well-known model: the main sequence. The deviation of saccades from this model has been reported in the case of various vision and neurological defects. However, we have observed large deviations from this model in saccades even in normally sighted subjects without any cognitive or neurological defects when viewing unconstrained visual stimuli such as movies. Dr. Pundlik is working on developing a model that explains the variance in the observed naturalistic eye movement data with the expectation that it will provide some insight into the neuronal processes involved in saccadic eye movement planning and execution.
14 (more than 1000 citations, Google Scholar as of August 2018)
Dr. Pundlik has published 22 peer-reviewed journal articles and conference papers. Below is a list of selected publications. View his publications on PubMed or Google Scholar.
- Pundlik S, Tomasi M, Moharrer M, Bowers A, Luo G. Preliminary Evaluation of a Wearable Camera-based Collision Warning Device for Blind Individuals. Optometry and Vision Science, 2018; 95(9): 747-756, 2018.
- Pundlik S, Yi H, Liu R, Peli E, Luo G. Magnifying Smartphone Screen using Google Glass for Low-Vision Users. IEEE Trans Neural Syst Rehabil Eng. 2016 Mar 23.
- Tomasi M, Pundlik SJ, Bowers A, Peli E, Luo G. Mobile Gaze Tracking System for Outdoor Walking Behavioral Studies. Journal of Vision. 2016;16(3):27.
- Pundlik SJ, Tomasi M, Luo G. Evaluation of a portable collision warning device for patients with peripheral vision loss in an obstacle course. Investigative Ophthalmology and Visual Science. 2015;57(4):2571-2579.
- Pundlik S, Woodard D, Birchfield S. Iris Segmentation in Non-Ideal Images Using Graph Cuts. Image and Vision Computing. 2010;28(12):1671-1681.
U.S. Patent Application 14/430, filed 07/2013; granted in 08/2017.
Computer algorithm and its implementation for predicting collisions by processing videos from a camera. The collision warning method implemented in the form of a wearable camera provides warnings to visually impaired people to prevent falls and collisions while walking. The same algorithm is capable to providing forward and lateral collision warnings in vehicles.
Mobile Device Application for Ocular Misalignment Measurement
U.S. Provisional Application No. 62/295,869, filed in 02/2016. Non-provisional patent application filed in 02/2017.
A smartphone app for objective measurement of strabismus angle. This app is intended for use in clinic and in home to rapidly measure eye alignment using a smartphone camera without using any additional attachments.