2017 IEEE International Conference on Image Processing

The 2017 IEEE International Conference on Image Processing (ICIP 2017) is fast approaching, and one of Kitware’s computer vision researchers, Dr. Zhaohui H. Sun, will be in attendance!  This event will be held in Beijing, China, September 17-20. It will “feature world-class speakers, tutorials, and industry sessions to create an excellent forum to foster innovation, entrepreneurship, and networking with academia and industry working in this field.”

Dr. Sun will present three papers in this selective international conference, sharing different techniques and novel methods for image processing.  During his presentation “Reflection Correspondence for Exposing Photograph Manipulation,” he will discuss utilizing image reflections as a strong indicator of content manipulation.  Image reflections are used because they are extremely difficult to fake, since they involve complicated interactions between surface materials, geometry and lighting.  The proposed technique should be able to identify images and video that have been maliciously manipulated. Authors include Eric Wengrowski with Rutgers University, Dr. Sun and Dr. Anthony Hoogs.

Dr. Sun will also present “Compact Image Representation by Binary Component Analysis,“ a method for fractional dimension reduction and face representation. Authors for this paper are Dr. Sun and Dr. Hoogs. The novelty of this paper is an improvement over principal component analysis (PCA) based methods by using binary component analysis (BCA) as an alternative.  Base vectors are restricted to binary values of +1 and -1, resulting in fractional dimension reduction and better face classification performance.

Finally, Dr. Sun will provide the presentation “Reconstruction of Highly Structured Image by Entropy Optimization.”  This presentation will detail an image reconstruction approach that proposes entropy optimization as an alternative to the L0 and L1 optimization that is so widely used in compressive sensing.  Dr. Sun will discuss two reconstruction algorithms based on this approach to address highly structured signals and images with multiple dominant modes.

Kitware’s computer vision team continuously delves into and applies computer vision and deep learning in order to develop and optimize innovative technologies that focuse on image and video processing.  These techniques help to build tools and capabilities to support government and commercial customers, as well as to collaborate with the academic community.  Focus areas such as image and video forensics, complex activity and threat detection, social multimedia analysis, super resolution and 3D vision all benefit from the research and development performed by the computer vision researchers at Kitware.  In addition, the open-source image and video analytics platform, the KitWare Image and Video Exploitation and Retrieval (KWIVER) toolkit, provides the image processing community with an advanced collection of computer vision tools without restrictions.

Please reach out to Dr. Sun at computervision@kitware.com to schedule time to discuss computer vision and deep learning approaches to image processing at ICIP 2017.