According to the latest news, a group of students of IIT Madras have created a new way to restore blurred photo. They published their paper in the IEEE Journal.
We like to click pictures and store them on our desktop or mobile as a memory. Unfortunately, all pictures are not perfect all the time. Sometimes, the images taken are good when they were captured, but they are ruined due to resizing or some other factor. It seems, now there is a new way that will solve the problem of blurred photos.
Researchers at the Indian Institute of Technology (IIT) Madras have come out with a new way to restore blurred photos. IIT Madras image processing and computer vision lab is headed by Dr. Rajagopalan. Here, the IIT team have used artificial neural networks to restore degraded images. The IIT team published their paper in IEEE Journal for selected topics in Signal Processing.
As mentioned earlier, in the paper, the group used a network of artificial neural groups to clean the blurred images. They have used the available database of environmental agents to test the efficiency of their model.
Dr. Rajagopalan said, “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. ”
Initially, it was difficult to identify the single neural network and clean the blurry parts but soon the team developed a two step system to address the issue.
The first step is degradation localization, where neural networks simply identified and removed the degraded parts of the photos. The second step is degrading Region Guided restoration where and to the image is cleared with the help of information collected from the first step. The purpose of the second step is to guide the restoration process. In the first step, one of the network layers performs a localization process and then transfers the information collected to the “Main restore network”.