2707

Content-Specific Image Enhancement

Lischinski Daniel, HUJI, School of Computer Science and Engineering, CS - Computer Vision

tamir_huberman_-_dani_lischinski1.jpg

Automatically enhances parts of image according to context

Categories

Computer Science & Engineering  Imaging / Computer Graphics

Development Stage

Research complete

Patent Status

Patent application filed in the United States

Highlights

  • Digital images captured by non-professional photographers requires effective automatic photo enhancement.
  • Most current systems manipulate tone and color over the whole image without understanding and accounting for the content of the image.

Our Innovation

Photo enhancement method that automatically improves the quality of specific image elements such as the sky, human faces, and underexposed salient regions by intelligently taking into account local and global semantics.

Figure 1: Comparison of original on left and results of new method on right

Figure 2: Input images (first left) are compared with our results (second), Photoshop Elements (third), Microsoft Office Picture Manager (fourth) and Google’s Picasa (fifth) 

Key Features

  • Intelligently combines a variety of local tone mapping and color correction operations while taking into account image content.
  • Using existing robust detectors of sky and people, image areas containing human faces, skin, blue sky and clouds, as well as salient underexposed parts of the scene are subjected to customized enhancement operators.
  • Consistently significantly improves the majority of images
  • Outperforms the automatic enhancement operators available in several popular products

Development Milestones

  • Seeking cooperation for licensing and commercialization

The Opportunity

  • Currently, 30 billion images a year are uploaded to Facebook
  • Automatic photograph enhancement tools are included in such popular commercial software packages as Adobe Photoshop, Microsoft Office Picture Manager, Google Picasa, and others.

 

Patent Status

Granted US 9,256,927

Contact for more information:

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