Researchers from IIT Madras have harnessed the power of artificial neural networks to restore CCTV images that have been degraded due to weather conditions, beyond recognition.?
Also Read:?IIT Madras Starts Online BSc Degree For 250,000 Students, Apply Till September 15
Published first in the journal IEEE, the technology is the brainchild of Dr A.N. Rajagopalan, holder of the Sterlite Technologies Chair Professor position in the Department of Electrical Engineering at IIT Madras and was assisted by Maitreya Suin and Kuldeep Purohit from IIT Madras.
Rajgopalan's Image Processing and Computer Vision Laboratory at the aforementioned institution is harnessing the power of artificial neural networks to bring these degraded images back to life. The new method at play can be utilised to clean images affected by rain-streaks, raindrops, haze and motion blur.?
Researchers discovered that looking at the degraded portions as well as cleansing the image was nearly impossible for a single neural network, so they decided to separate the tasks into two different stages.?
In the primary stage, one neural network was trained to localise the degraded part whereas in the second stage, the neural network uses this very information to restore the image. Dr Rajagopalan explains, “Bad weather in the form of rain and haze causes significant loss of image quality.?
Also Read:?IIT Madras Scientist Grows Tiny Brains Using 3D Printing For Brain Research
The presence of raindrops on camera lenses is a related problem that poses its own set of challenges. These effects not only impact human visibility but can also adversely affect the performance of computer vision systems meant for autonomous driving, drone imaging, and surveillance, to name a few.?
These degradations have high spatial variability due to non-uniform depth variations in the haze, drop sizes and their locations in raindrops, and rain streak directions and locations.”For the neural networks, the team made use of publicly available datasets of rain streak, haze, raindrop and motion blur to test their model.?
They applied knowledge distillation for image restoration along with the prediction of degraded locations. This allowed the system to outperform existing competing methods.
Also Read:?IIT Madras Offers World's Cheapest Data Science Course That Any Student Can Learn For Rs 1,000
He added, “Our premise is to use the auxiliary task of degradation mask prediction to guide the restoration process. We demonstrate that solving this auxiliary task injects crucial localizing ability in network layers. We transfer this ability to the main restoration network using attentive knowledge-distillation and focus on the refinement of degraded regions by exploiting this additional knowledge.”?
Even though this tech is only targeted towards restoring degraded CCTV imagery, it would be really cool to see if this tech can be applied for restoring physical images that have aged poorly over the past, and maybe help us relive those memories.