Research

The usual paradigm of computational photography is to jointly design hardware and software for specific applications to enhance overall performance and efficiency. However, more often than not, devising new hardware for each application is cumbersome and expensive. These hardware setups are often bulky and require meticulous calibration, making them ill-suited for practical, real-world scenarios. My research tackles the well-established computational photography problems and achieves better performance by re-purposing existing hardware with minimal changes. While this seems ambitious, my research uses state-of-the-art deep learning techniques to achieve this goal.

While deep-learning algorithms achieve significantly higher performance on several tasks, they still require informative inputs. Images and videos captured using traditional cameras have their limitations and are not always suitable for several applications. For instance, a learning-based algorithm proposed for video reconstruction from a single blurred image suffers from motion ambiguity leading to poor reconstructions. An additional input that encodes the motion information would significantly improve the algorithm performance. Hence, in my research, I go beyond the traditional cameras and explore novel sensors and setups most suitable for my application. Along with the suitability, aspects like the hardware's power consumption, commercial availability, form factor are also considered. After choosing the proper hardware, the task is to design a novel learning-based algorithm for each application. My research considers two well-studied computational photography problems: high-frame-rate video and light-field video reconstruction.

Select Publications

Thesis
  • Prasan Shedligeri (2022), "Reconstructing High Temporal and Angular Resolution Videos from Low Data Bandwidth Measurements", PhD thesis submitted to Indian Institute of Technology Madras, India. [PDF]
Published Papers
  • Shrisudhan G, Prasan Shedligeri, Sarah, Kaushik Mitra (2022), "Synthesizing Light Field Video from Monocular Video", Accepted at European Conference on Computer Vision (ECCV) 2022. [Preprint] [Code] [Webpage].
  • Prasan Shedligeri, Florian Schiffers, Sushobhan Ghosh, Oliver Cossairt & Kaushik Mitra. (2021), "SeLFVi: Self-supervised Light Field Video Reconstruction from Stereo Video", Accepted at IEEE International Conference on Computer Vision (ICCV), [Preprint], [Supplementary pdf], [Webpage], [Code].
  • Prasan Shedligeri, Florian Schiffers, Semih Barutcu, Pablo Ruiz, Aggelos Katsaggelos & Oliver Cossairt. (2021), "Improving Acquisition Speed of X-Ray Ptychography through Spatial Undersampling", Accepted at IEEE International Conference on Image Processing, [Preprint].
  • Prasan Shedligeri, Florian Schiffers, Semih Barutcu, Pablo Ruiz, Aggelos Katsaggelos & Oliver Cossairt. (2021),"Regularization for Undersampled Ptychography", Accepted at OSA Computational Optical Sensing and Imaging (COSI) 2021, [Preprint].
  • Prasan Shedligeri & Kaushik Mitra. (2021), "High Frame Rate Optical Flow Estimation from Event Sensors via Intensity Estimation", Elsevier Journal of Computer Vision and Image Understanding, 208-209, doi:10.1016/j.cviu.2021.103208 [Preprint] [Webpage]
  • Prasan Shedligeri, Anupama S & Kaushik Mitra. (2021) CodedRecon: Video reconstruction for coded exposure imaging techniques. Accepted at Elsevier Journal of Software Impacts, doi:10.1016/j.simpa.2021.100064 [Paper] [Code] [Webpage]
  • Prasan Shedligeri, Anupama S & Kaushik Mitra. (2021), "A Unified Framework for Compressive Video Recovery from Coded Exposure Techniques", IEEE/CVF Winter Conference on Applications of Computer Vision, doi: 10.1109/WACV48630.2021.00164 [Preprint] [Slides] [Supplementary] [Code] [Webpage]
  • Anupama S, Prasan Shedligeri, Abhishek Pal & Kaushik Mitra. (2020), "Video Reconstruction by Spatio-Temporal Fusion of Blurred-Coded Image Pair", IEEE International Conference on Pattern Recognition, doi:10.1109/ICPR48806.2021.9412968 [Preprint] [Slides] [Supplementary] [Code] [Webpage]
  • Prasan Shedligeri & Kaushik Mitra. (2019), "Photorealistic image reconstruction from hybrid intensity and event-based sensor", Journal of Electronic Imaging, 28(6), International Society for Optics and Photonics. doi:10.1117/1.JEI.28.6.063012. [Paper] [Arxiv preprint]
  • Prasan Shedligeri & Kaushik Mitra. (2019) "Live Demonstration: Joint Estimation of Optical Flow and Intensity Image From Event Sensors", IEEE CVPR workshop on Event Sensors. [Paper]
  • Prasan Shedligeri, Sreyas Mohan & Kaushik Mitra. (2017) "Data driven coded aperture design for depth recovery", IEEE International Conference on Image Processing, 56-60. doi:10.1109/ICIP.2017.8296242 [Paper] [Preprint] [Website]