3005

Dehazing using Color-Lines

Fattal Raanan, HUJI, School of Computer Science and Engineering, CS - Computer Vision

Color-lines used to recover scene transmission based on offset of lines from the origin

tamir_huberman_raanan_fattal2.jpg

Categories

Computer Science & Engineering, Imaging / Computer Graphics

Development Stage

Research completed, method implemented and tested against
state-of-the-art

Patent Status

US patent application filed

Highlights

  • Most image dehazing methods remove the layer of haze by recovering the direct scene radiance.
  • These methods rely on a physical image formation model that describes the hazy image as a combination between the scene radiance and the atmospheric light.
  • In the case of RGB images, the input image supplies only three of the four constraints required to produce optimum results and necessitate supplying additional information about the scene.
  • New method produces superior results without the need for additional inputs.

Our Innovation

New method for single-image dehazing that relies on a generic regularity in natural images

where pixels of small image patches typically exhibit a one-dimensional distribution in RGB color space, known as color-lines. Local image formation model uses the color-lines in hazy images to recover the scene transmission based on their offset from the origin.

 

Key Features

  • Algorithm validates its hypothesis and obtains more reliable transmission estimates where possible.
  • Unique pixel distribution predicted by model enables identification of patches not exhibiting proper color-lines and discards them.
  • Extensive evaluation of this method over different types of images and comparison to state-of-the-art methods over established benchmark images shows consistent improvement in the accuracy of the estimated scene transmission and recovered haze-free radiances.

Development Milestones

  • Real-time implementation on various platforms
  • Developing application specific user-interface

The Opportunity

  • Applications in long-distance photography
  • Airborne imaging
  • Satellite photography
  • Underwater photography
  • Security/Surveillance

Contact for more information:

Eitan Dekel
VP Business Development - Computer Science
+972-2-6586692
Contact ME:
Image CAPTCHA