Log on:
Powered by Elgg

Christopher Holder :: Blog :: Proposal!

April 18, 2008

Student Name: Chris Holder and Paula Carrillo


Supervisor’s Name and Title at FIU/FAU: Hari Kalva, Assistant Professor, FAU


Name of the PIRE International Partner’s Institution: Tshingua University


Supervisor’s Name and Title at the PIRE International Partner’s Institution:

 Shiqiang Yang, Professor, Dept. of Computer Science and Technology


Project Title: Data Mining for Video Transcoding and Encoding


Problem Statement: This project aims to exploit structure in video to enable low complexity encoding. The proposed methods use data mining to reduce the complexity. The goal is to develop data mining tools to predict complex coding modes in videos. The tools exploit structural similarity in video. One of the objectives is to develop a minimum set of attributes to make optimal decisions. The methodology developed will be useful in other data mining problems.   


Motivation and Impact: Data mining has been applied in video analysis and understanding but there has been little work in improving encoding performance using data mining. The proposed approach of exploiting structure in video allows more efficient encoders. The proposed solution improves encoder parallelization. The proposed work will enable better utilization of multi-core and upcoming many-core systems. The proposed project will have applications in high-density video processing systems.


Current Status: This work is part of ongoing research in the Multimedia Lab at FAU. Dr. Yang’s group at Tshingua is also working on advanced video encoding algorithms. We expect to explore the complexity vs. quality tradeoffs. Preliminary results confirm the potential of data mining in video coding applications.


Research Roadmap:

Jan 15 to May 15 2008 – Continuing work on applying data mining for H.264 encoding and transcoding

Mar 15 to May 15 2008 – Exchange of ideas with Prof. Yang and developing a research plan for the students’ stay in Beijing

May 15 – Jul 15 2008 – Research plan execution. Development of data mining techniques for video coding.

Aug 15 – Dec 15 2008 – Completion of simulation, analysis, and paper preparation

Paper Submission:

Early July submission to IEEE International Conference on Consumer Electronics (ICCE 2009)

December submission to IEEE International Conference on Image Processing (ICIP 2009)


Relation to PIRE Core Research Projects: The data mining tools developed fit into  the CI Integration layer. The video coding application fits into CI Application layer. Low complexity video coding has applications in healthcare communication and visual data acquisitions in hurricane mitigation.

Posted by Christopher Holder

You must be logged in to post a comment.