Automatic Photo Enhancement Server (HIPIE 2)

HIPIE 2 is an internal name for the HP Smartstream Photo Enhancement Server, robust, scalable, and automatic photo image enhancement software, customized for photo specialty workflows fulfilled using Indigo presses. It is designed for 24/7 operation, without human intervention, and is part of HP Indigo's Smartstream workflow offering. The codename HIPIE stands for the original "HP Indigo Photo Image Enhancement" name, with HIPIE 2 being its second version.

With the advent of digital press, photo specialty prints emerged as an important category for commercial print. This includes photobooks, customized posters, and calendars. These applications include consumer photographs, which are not as optimized for print as those professional photographs used in traditional jobs. Consumer pictures are often too dark or too bright, not colorful enough, and/or not sharp enough, for instance. As a result, Print Service Providers that fulfill photo specialty workflows have the need to improve image quality in addition to keeping a high print quality.

HIPIE 2 is an automatic image enhancement package, which works as part of a commercial print environment, enhancing each incoming photograph in a fast and reliable way. It improves the visual appearance of the photos as they are printed in photo specialty products, without human intervention. It is commercialized by HP Indigo as part of their photo quality offering, and as part of the HP Smartstream workflow offering, under the name "HP Smartstream Photo Enhancement Server".

  Image Enhancement
HIPIE 2's image enhancement is state-of-the-art. We have introduced innovative enhancements, such as 3D boosting and face make-up. Image analysis modules determine the amount and type of enhancements needed for each individual photograph. The user is able to apply the following image processing tasks. All these tasks are completely automatic, but their strengths can be influenced by the user controls.
  • Sharpening Each input image is sharpened as needed, so that it appears at the right sharpness in print.
  • Noise reduction Noise is estimated, and then reduced, without damaging texture or fine details in the image.
  • 3D boosting This is a novel feature, whereby shadings are enhanced so to give an impression of increased depth to the objects in the image (see Fig. 2).
  • Contrast enhancement Low-contrasted images have their effective dynamic-range stretched; this includes lowexposure images, which are brightened (see Fig. 3).
  • Shadow detailing HIPIE 2 detects the low-exposed regions of the image, and brightens them, without demaging the well-exposed regions of the image (see Fig. 4).
  • Face make-up Another novel feature, this makes sure that skin on faces are smoothed, as if the subject applied makeup, reducing the visibility of wrinkles and minor blemishes.
  • Chroma boosting HIPIE 2 enhances overall image chromaticity.
  • JPEG artifact reduction Blocking and ringing artifacts, typical from strong JPEG compression, are reduced.
  • Face color correction Skin color is analyzed and corrected for photos with a color cast.
  • Resolution Enhancement If the image resolution is too low, compared to the target print size, then the resolution is doubled by a smart interpolation algorithm in order to avoid jagginess or pixelization on edges.
  • Red-eye removal Optionally, HIPIE contains a module for automatically detecting and removing red-eye effects.
  Example Enhancements
Face make-up
Skin color correction
3D Boosting
  System Infrastructure

HIPIE 2 is intended to fit seamlessly into high-throughput photo specialty workflows, which has a number of implications. First, it must be robust, meaning that it must be able to run nonstop for months on end and it must be able to recover automatically from any possible failures. Second, it must be scalable, meaning that it must be able to utilize any number of processors in a distributed computing environment. Third, it must be easy to configure and manage.

HIPIE 2 uses a master-slave architecture and has five components. The Windows service is used to (automatically) start and stop the other components of the system, and there is a Windows service component on each host in the cluster. Hot folder processes are responsible for 'watching' each configured hot folder, and ensuring that each input file is processed appropriately by a worker process. The queue process matches work requests from hot folder processes with work offers from worker processes, thus distributing work from the various hot folders throughout the cluster to worker processes. The worker processes simply request work from the queue process, enhance the image using the options passed as part of the request, and return the completion status to the requesting hot folder when done. A logger process per host in the cluster is used to record log messages from the various components in the system.

  Publication
  • Automatic Photo Enhancement Server (HIPIE 2)
    Renato Keshet, Carl Staelin, Boris Oicherman, Mani Fischer, Pavel Kisilev, Sagi Schein, Doron Shaked, Marie Vans, Hila Nachlieli, Ruth Bergman, Ron Maurer, Michal Aharon, Alex Berkovich, Shimrit Feldman, Ran Waidman, Oren Bengi, Gal Amit, Shlomo Harush, Alon Asher, Matthew Gaubatz, Steven J Simske, Darryl Greig, Hui Chao, International Symposium on Technologies for Digital Fulfillment (TDPF), 2009.
       
  • Perceptual Segmentation: Combining Image Segmentation With Object Tagging
    Bergman, R.; Nachlieli, H. (2011). IEEE Transactions on Image Processing (June 1, 2011), 20(6), 1668-1681. DOI: 10.1109/TIP.2010.2088970
     
  • HP Indigo Photo Enhancement Server
    Product Description, 2007.