When taking a photograph, there are many reasons why the images may present visibility distortions. Even many current computer vision methods based on machine learning architectures do not work when the input image is affected by climatic distortions or environmental conditions such as fog, rain or dust. In addition, images can be captured in poor visibility conditions, such as at night or under water. The images obtained in these circumstances often have faded colours and contrast loss, among many other problems.

The aim of the project In-Camera Visibility Restoration (ICVIR) is to develop new image processing technology integrated into a digital camera (DSLR cameras or mobile phones) that, given an input image, decides whether there is visibility distortion and, if necessary, applies the necessary image processing techniques to allow the camera to restore any loss of visibility caused by uneven conditions, allowing obtaining the restored image at the actual time of capture.

Javier Vázquez-Corral, a member of the Image Processing for Enhanced Cinematography research group (IP4EC) at the UPF Department of Information and Communication Technologies (DTIC), is the principal investigator of the ICVIR project, with a duration of three years, funded by the Spanish Ministry of Science, Innovation and Universities with a budget of 181,500 euros.

Removing poor visibility conditions will be useful for surveillance cameras and driverless smart vehicles and drones

As Javier Vázquez, principal investigator of the ICVIR project points out, removing visibility shortcomings from images is useful not only for aesthetic purposes but also for security reasons. “For aesthetic reasons, every amateur photographer likes to get the best picture possible at the time of image capture, without the need for further editing”.

“With regard to the usefulness of this new technology for security reasons, the removal of poor visibility conditions will be useful for surveillance cameras and smart vehicles (driverless vehicles and drones, for example) as it allows enhancing the performance of computer vision systems that need to operate in the open air under uncontrolled conditions”, Vázquez explains.

This technology has many more fields of application: in archaeology and underwater pipeline inspection, for images of astronomy, in medical imaging, etc.

Digital cameras can be fully functional even in poor visibility conditions and will avoid the need for further image processing

In short, this project will allow digital cameras to be fully functional even in poor visibility conditions and will avoid the need for further image processing.

Innovation must meet efficiency requirements

For a manufacturer to incorporate an innovation to a digital camera, such as a new image processing algorithm, this innovation must demonstrate that it  fulfils a series of requirements in terms of quality, robustness, computational complexity and execution time. “This project will ensure the integration of new algorithms into digital cameras in terms of quality and efficiency”, Vázquez comments.

Project title and acronym:

In-Camera VIsibility Restoration (ICVIR)

Principal Investigator: Javier Vázquez Corral

Budget: 181,500 euros.

Duration: 3 years