Efficient and Quality Contouring algorithms on the GPU, in Computer Graphics Forums 2010.

MEDICS: ultra-portable processing for medical image reconstruction, from Umich agian, in PACT 2010. An excellent example for domain specific architecture. It is unclear how the Hough transform is mapped onto the MEDICS.

Some interesting articles from CACM Oct 2010:

• A neuromorphic approach to computer vision, with illustrative figures.
• Photoshop scalability: keeping it simple, how photoshop goes parallel.
• Thinking clearly about performance, part 2, part 2 of the “Oracle Tuning”.
• Platforms and services: understanding the resurgence of Apple, the business model behind Apple.
• SCORE: agile research group management, some funny stories about managing a research group.
• Should code be released?. Yes, in short.
• Tackling architectural complexity with modeling, not specific to hardware architecture.

The taster of MICRO 2010. The titles alone are enough to get me excited. Again, multi-core and many-core are hot topics of the conference.

Toward a multicore architecture for real-time ray-tracing, in microarchitecture 2008. An epic example of domain specific architecture exploration.

An instruction-systolic programmable shader architecture for multi-threaded 3D graphics processing, in j.jpdc 2010. It proposed systolic execution of warps. The handling of branch is different than Wilson Fung’s dynamic warp formation. It is unclear how it handles the shared memory bank conflicts and global memory coalescing problem, if it has a memory hierarchy like NVIDIA’s. Anyway, the analysis of this paper is awesome, especially the analysis on the vector width.

FPGA Implementation of Parallel Histogram Computation, from Delft. Literatures before 2007 are nicely summarized.

Good features to track, by Shi and Tomasi, in CVPR 1994. It is used by OpenCV.

A Comparison Study on Implementing Optical Flow and Digital Communications on FPGAs and GPUs, from CHREC at BYU, in TRETS 2010.

The Middlebury Computer Vision Pages contains large dataset for stereo vision and optical flow.

Characterizing the impact of process variation on 45nm NoC-based CMPs, in j.jpdc 2010. (non-DOI link)

Cache-Based Memory Copy Hardware Accelerator for Multicore Systems, by Filipa Duarte from the Ultra Low Power DSP group of IMEC/Holst Center Eindhoven and Stephan Wong from Delft, in TC 2010. Both analytical model and simulation are used.

Design Space Exploration for Efficient Resource Utilization in Coarse-Grained Reconfigurable Architecture, from Samsung, TAMU, and SNU, in TVLSI 2010.

The micro-architecture of Sandy Bridge and Bulldozer, by David Kanter from realworldtech.

Two similar surveys on optical flow:

• Performance of optical flow techniques , in International Journal of Computer Vision 1994.
• The computation of optical flow, in ACM Computing Surveys 1995.

Stability of networked control systems, in Control System Magazine 2001.

Proceeding of IEEE has a special issue on networked control systems in 2007.

Exascale Computing Study Report, a comprehensive overview and prediction of computing. The figures in the report are inspiring. The embedded computing community should come up with a report like that. The HiPEAC vision report is not as illustrative.

Motion algorithms used in OpenCV:

• The representation and recognition of human movement using temporal templates, in CVPR 1997.
• Motion segmentation and pose recognition with motion history gradients , in Machine Vision and Applications 2001.

Optical flow algorithms used in OpenCV:

• Determining optical flow, by Horn and Schunck, in Artificial Intelligence 1981.
• An Iterative Image Registration Technique with an Application to Stereo Vision, by Lucas and Kanade, in 1981.

Off topic:
Some interesting articles on modelling:

• Models in Science, in Stanford Encyclopedia of Philosophy.
• The Scope and Method of Economics, in The Review of Economic Studies 1945.
• The Financial Crisis and the Systemic Failure of Academic Economics , a Univ. of Copenhagen Dept. of Economics Discussion Paper published in 2009.

The Future of Computer Architecture Research, post by CCC. Report of the Advancing Computer Architecture Research (ACAR) workshop is also available. The motivation of Domain Specific Languages (DSLs) is well written.

Compiler research: the next 50 years, by Mary Hall, David Padua, and Keshav Pingali, in CACM 2009.

Intel Array Building Blocks (ArBB), formerly known as Ct, was released.

CenSurE: Center Surround Extremas for Realtime Feature Detection and Matching , by Motilal Agrawal, Kurt Konolige and Morten Rufus Blas, in ECCV 2008. It is used in the star detector of OpenCV.

A NoC-Based Hybrid Message-Passing/Shared-Memory Approach to CMP design, in j.micpro 2010. The first work to address the implementation issues of hybrid memory model in NOC. (non-DOI link)

Machine Vision Applications (MVA) 2011 calls for paper. Deadline is on December 15, 2010.

ParCo 2011 calls for paper. Deadline is on 20 March 2011.

Off topic:
“Companies don’t go out of business because they run out of cash,” Jen-Hsun said in a fireside chat, “The reason most companies go out of business is they run out of ideas.”

Image inpainting resources from the web page of marcelo bertalmío and Guillermo Sapiro.