Real Innovation@Xerox: Xerox Researchers Crack the Code on Easier Ways to Cut Through the Clutter of Documents and Photos; Xerox Press Release
Say you are a graphic designer working for a global health organization. You need to find the right photograph that conveys how critical good medical practices are to the lives of children. Using today’s imaging technology to sift through large photo collections and identify that image could take hours.
Today Xerox researchers have released technologies that make finding images to illustrate and communicate your message, easy. The imaging tools are available in the company’s online technology portal called Open Xerox, a web site that allows customers to demo new inventions and provide feedback.
The most recent crop of technologies, developed at the Xerox Research Centre Europe, are designed to help customers manage overflowing collections of information, documents and images. The new tools use machine learning and other techniques to analyze, classify and retrieve images and collections in new, creative ways. For example, one tool can sort vast photo collections according to a particular aesthetic style instead of a basic category.
"The demos illustrate how our imaging technology could be used to solve real business problems,"said Craig Saunders, manager of the computer vision research group based in Europe. "These technologies help customers find exactly what they need among huge collections of photos, scanned documents, drawings or marketing materials."
By making technology available through Open Xerox, the company invites customers, partners, technology experts and early adopters to play with the demos. "We want to know what you think, so we can continue to refine and test our technology to meet real needs,"said Saunders.
The new imaging demos available are:
- Similar Image Search – Capable of searching five million images in less than a second on a standard desktop computer, this technology efficiently allows users to search large collections of high quality photos, digital assets, graphic design elements, scanned documents or other media to help pinpoint the exact item you need. By representing images in a smart way (using robust, efficient, compressed image signatures) the technology works with all types of diverse media, providing a versatile search tool that does not rely on the cloud and is optimized for efficient memory and processing The technology compresses all the digital media it finds providing a critical building block for future services and applications The technology behind Similar Image Search can be used to help organize, visualize and query all types of image based repositories, from large photo stock databases, enterprise scale digital document repositories or a personal photo collection on your desktop.
- Image Categorizer – Previously announced, this technology is now available for testing and automatically classifies images according to their content. Targeted at advertising, marketing or photographers who manage large collections of images, it uses machine learning techniques and image processing algorithms to analyze and classify images into almost 700 different categories. Being able to classify images only on their visual content enables the automation and improvement of many tasks. For example, textual queries can be made without having to manually add relevant tags for each image. The technology also allows automated workflow routing based on image content alone often eliminating the need for costly Optical Character Recognition.
- Aesthetic Image Search – This new alpha prototype demo makes use of the image categorizer, but goes one step beyond and tackles the difficult task of trying to learn what makes an image special and unique. Many current methods for image classification are solely based on recognition of parts -- if you find some wheels and a road, then the picture is more likely to contain a car than a giraffe. In order to determine the aesthetic quality or stylistic elements of an image, a new approach is needed. Applications include searching large photo databases; helping to select the best photos for archival or printing or for marketing or advertising agencies that want to select a particular style of image among a vast collection.
- Catepix – Sophisticated image analysis technology is put to use in this fun application that will analyze your FacebookTM photos and determine how popular your images are among your Facebook friends and what your photos say about your personality. Add it to your profile and see what it says about you!
About Open Xerox
Open Xerox, launched last year, was co-created by researchers at the "Xerox Innovation Group and PARC, the Palo Alto Research Center. Users not only can try the services and applications online but also register to access a wider range of technologies from early stage alpha projects to more mature technologies, and can run services and evaluate technologies by integrating them directly into their own applications. Researchers use Open Xerox to gather feedback on new research prototypes as well as mature technologies.
Xerox Press Release | Tuesday 15 November 2011
Read the article