IIT Madras have found a way to restore fuzzy photos

Indian Institute of Technology (IIT) Madras has found a way to restore fuzzy photos

Source from:- Indian Institute of Technology (IIT) Madras

We’ve all been there before: Rather cool is happening and you want to take a picture of it. You switch out your phone, take the picture only to realize that the photo is fuzzy and unclear. You try to take an alternative shot but that moment has passed, never to be captured again.

I’ve personally been in that situation before, so I know full well how maddening that can be.

Everyone likes to plosive pictures and accumulation them to create long memories. Notwithstanding, not all of the pictures uprise out perfect all clip. The images that are stored on mobile phones and laptops await splendid when it was captured, but due to environmental push, it gets smash. Specified problems are not heard, but now there is a way to spend photos digitally and modify the fuzzy photos.

Researchers at the Indian Institute of Technology (IIT) Madras has found a way to restore fuzzy photos. Dr. Rajagopalan is leading the IIT Madras image handling and computer vision lab. They are using artificial neural networks to reinstate the degraded images.

 

In latest times, the IIT team has published their work in IEEE Journal for selected topics in Signal Handling. It shows the techniques which are developed by them. The technique is to use a network of artificial neural groups to clean the images that are fuzzy due to raindrops, rain streaks, motion blur, etc. For their model, the team has used the existing database of environmental agents.

 

The arena of photography has come a long way from being an implement to preserve memories to being used now for observation, drone flying, futuristic self-sufficient driving systems and various other domains. These applications require that the images are clean and not fuzzy or impaired. However, weather conditions pose a major challenge in getting clean images. In places such as mountainous areas or where rainfall is common, surveillance cameras that are placed in cities cannot be relied upon for watching as they cannot provide clean camera footage.  

Dr. Rajagopalan described the study:

 

“Bad weather in the form of rain or haze causes significant degradation in image quality. The presence of raindrops on the camera lens is a related issue that poses a series of challenges in itself. Not only does it affect human vision, it can also adversely affect the performance of computer vision systems intended for automated driving, drone imaging, and surveillance. These degradations result in uneven haze depth. Spatial variability is greater due to variability, droplet size and position within the raindrop, and the direction and position of the rain streaks. ”

 

At the time of investigation, it was difficult to identify the single neural network and clean the fuzzy parts on the images. Then they made the system in which the process took place in two different stages. The first step is called degradation localization, neural networks worked to identify and remove the tainted parts of the photos.

The second step is to degrade Region-Guided restoration and to clear the image, the information that is provided in the first step is used. The main resolution is to guide the restoration process.

One of the network layers in the first step make a localization process and then it transfers the information collected to the “Main restore network”.