Researchers from the Indian Institute of Technology Madras (IITM) and Rice University (USA) have developed algorithms for Lensless Miniature Cameras (LMC). Such lensless cameras have multiple vision based applications in augmented reality, Virtual Reality, Safety, Intelligent Wearables and Robotics. There is certain limitation such as price, form-factor, and weight.

Lensless cameras don’t have a lens as in a traditional camera but functions as a working element allowing the detector to catch a sharp picture. IITM and Rice University scientists have developed a deep learning algorithm for generating photorealistic images in blurry lensless capture. Taking out a lens may result in the miniaturization of a camera. Researchers internationally are searching for replacements for such lenses, stated IITM.

In 2016, Prof. Ashok Veeraraghavan’s lab at Rice University was successful in producing a lensless camera. They could come up with a cheap and low-weight ultra-thin lensless camera. The operation of lenses is to focus on the incoming light. In these recently developed lensless cameras, a thin optical mask has been placed just before the detector with a distance of roughly 1mm. But due to the lack of focusing components, the lensless camera captures fuzzy images restricting their usage. Due to the absence of a focusing element, the lensless camera captures a multiplexed or globally blurred measurement of the scene.

Scientists have now developed a computational solution for this issue. The group developed a deblurring algorithm that can fix the blurry pictures taken from a lensless camera. The findings have been presented as a paper in the prestigious IEEE International Conference on Computer Vision and an advanced version of pattern analysis and machine intelligence.

This research was directed at IITM by Dr Kaushik Mitra, Assistant Professor, Department of Electrical Engineering. The study team comprised Salman Siddique Khan, Varun Sundar and Adarsh VR from IITM. Prof. Ashok Veeraghavan headed Rice University’s group that consists of Dr Vivek Boominathan and Mr Jasper Tan.

Dr Kaushik Mitra further explains that existing algorithms to deblur images predicated on conventional optimization strategies yield low-resolution and noisy pictures. Our Research team employed deep learning to create a reconstruction algorithm referred to as ‘FlatNet‘ to get lensless cameras that led to significant improvement over conventional optimization-based algorithms. FlatNet was analyzed on various real and ambitious scenarios and discovered to work in deblurring images recorded from the lensless camera.

Further, Dr Mitra explained that lensless imaging is a new technology, and its real potential is in resolving imaging and vision issues. The technology hasn’t yet been exploited entirely. Hence we’re focusing on designing better and newer lensless cameras employing data-driven methods, devising efficient algorithms for performing inference on lensless catches and looking into intriguing and significant software like endoscopy and intelligent surveillance, among other regions where you can fully comprehend the advantages of lensless imaging.

This study was funded by National Science Foundation Career and NSF Expeditions (USA), Neural Engineering System Design (NESD), Defense Advanced Research Projects Agency (DARPA), National Institutes of Health Grant, along with Qualcomm Innovation Fellowship India 2020. News SourceIndia Science Wire