ICME 2016 Grand Challenge: Light-Field Image Compression

July 11th – 15th, 2016, Seattle, USA

Call for Proposals

The detailed information related to Light-Field Image Compression Grand Challenge can be found in the Call for proposal document below.

Proposals Submission

Submission of the proposals is doen via FTP server. Unique access information will be given to proponents based on their request sent to organizers email.

Plese send the access request to martin.rerabek@epfl.ch

Challenge Description

Light field images capture information about the intensity of light in a scene and also information about the direction of the light rays in space. They are acquired with a light field camera typically corresponding to an array of micro-lenses placed in front of an otherwise conventional image sensor. In recent years, light-field cameras have become commercially available. However, an efficient compression format for this new modality does not exist yet. This challenge solicits contributions that demonstrate efficient compression of light field image data. Moreover, new evaluation methodologies are sought. Furthermore, additional publicly accessible light-field images along with evidence for compression efficiency as well as other attractive features are also accepted.

Dataset/APIs/Library URL

http://mmspg.epfl.ch/EPFL-light-field-image-dataset
The light-field image dataset contain images in LFR, as provided by the Lytro ILLUM camera, accompanied by their thumbnails, corresponding depth maps and relative depth of fields coordinates. It also contains Lytro camera calibrating data. A subset of images in this database will be used.

Evaluation Criteria

1. Evaluation of Compression Efficiency

Each image and any associated data used for the reconstruction and interaction with the light-field data shall be compressed at several bit rates on a selected subset of images from the above mentioned database (a detailed test procedure will be announced and communicated). The performance will be evaluated using conventional objective metrics (e.g. PSNR, SSIM, etc.) relative to the uncompressed image data at either multiple focal or view points, known only to the evaluators. Reference software to perform the rendering and manipulation functions will also be provided.

2. Verification of Evaluation Methodology

Proposals for new methodologies to evaluate compression of light field image data may demonstrate any one or several of the following:

  • Objective quality measures that are well correlated with subjective assessment of image quality over a range of manipulations
  • Methods that assess the quality of interaction with the compressed media such as assessment of free-navigation, etc.
  • Methods to assess efficiency of features such as change of focal point by means of blur metrics on regions of interest, etc.

Furthermore, proposers of such additional evaluation methodologies should provide a very detailed description along with an executable or a source code (preferred option) of the methodology they submit.

3. Additional content

Proposals for new content are welcome along with proposals for image compression and evaluation methodologies. Such content should be made freely available for research and standardization purposes.

 

Submission Guidelines

Submissions of light field compression algorithms or evaluation methodologies should provide a detailed technical description in the form of a short paper as well as material to validate the performance of the submission (e.g., compressed files and binary executable to reconstruct and measure performance). Submissions for additional content should be made along with either a compression algorithm or an evaluation methodology. The paper could be included as part of the ICME proceedings and published on IEEE Xplore if accepted after peer review.

Additional Information

A winner in each category will be selected by a judging committee. Specifics on the evaluation criteria and form of the award will be determined later. Financial sponsorship is being sought.
This challenge is aligned with the objectives of the new JPEG PLENO initiative. As such, the JPEG committee and its members are likely to support this challenge, e.g., helping to set target rates, providing anchor encodings to be used as a benchmark, assistance with evaluation of proposals, etc. COST Action IC1003 Qualinet members are also likely to assist in assessment of subjective and objective quality.

Questions and requests

If you have any questions or requests, or need further clarifications, please contact any of the organizers below:

Touradj Ebrahimi (EPFL)

Peter Schelkens (VUB/iMinds)

Fernando Pereira (IST)