A lot of my time is spent researching other work. Fortunately, as Computer Engineering/Computer Science is a relatively young field, a wealth of knowledge is available online. Like most universities, the library at FAU provides access to many online repositories. ACM and IEEE members also have access to their respective resources.

I usually search for a PDF version of a publication. A BibTeX citation is also convenient. A PDF of a publication can be saved and searched (very useful). If final, it is the authoritative version. Sometimes only the pre-publication manuscript is available. In this case be careful – changes and corrections may have occurred before publication.

If you are writing a publication you may be using LaTeX. BibTeX entries can easily (and properly) be included. I’ve been building a BibTeX database with references to hundreds of publications related to my dissertation which had been extremely useful.

IEEE Xplore

IEEE Xplore is a service for finding IEEE publications. PDF versions of most documents in its database are available. However, references and citing documents can be missing or incomplete. References are not available in BibTeX format. Still, IEEE Xplore often is a good place to start.

ACM Portal

ACM Portal offers the ability to search within “The Guide to Computing Literature”. This is a database of publications from the ACM and over 3000 other publishers. Full PDFs are available for ACM publications but missing for other sources. Like IEEE Xplore, references and citing documents are provided but sometimes incomplete. ACM Portal does provide BibTeX citations.

DBLP

DBLP is an excellent bibliography of Computer Science publications. BibTeX entries are provided for all publications in its database. Links to electronic editions of some publications are included as well, but mostly for more newer entries. The biggest shortcoming of DBLP is its incompleteness. Many publications have not yet been indexed by the service.

CiteSeer

CiteSeer is a broad catalog of scientific publications (not just from CS and CE). You may obtain more irrelevant results. Because CiteSeer can be edited by anyone it has become one of the largest repositories of publications. Unfortunately, this can also lead to errors or inaccuracies. Links to citing documents are references are provided. Other statistics, such as citations per year are also available. Also, BibTeX entries are provided, although I frequently encounter syntax errors. PDF versions of documents can be downloaded.

Google Scholar

The strength of Google Scholar is its breadth. It is an index of publications across all fields. While this distracts from a search for papers on focused topics (IEEE Xplore and ACM Portal are best for this) it is one of the best sources for finding citing documents. BibTeX entries are not provided. Additionally, links to electronic editions are inconsistent with some going directly to PDFs and other to the publisher’s site (where the publication can be purchased or downloaded if your University is a subscriber).

Library

Older publications can be among the most difficult to find. A library’s electronic collection is a useful resource for finding these papers if you know the journal and volume where they appeared. For example, the FAU Libraries Electronic Collection indexes thousands of databases.

Author websites

Many authors provide links and, sometimes, BibTeX entries for their publications. I’m in the process of cataloging all my publications on my research page. My dissertation advisor has been doing the same for years.

Proxy

As mentioned, some of the various resources redirect to the publisher’s page that provides a link to purchase the publication. This can happen if you are outside your university’s network and, as a result, cannot be authenticated as a subscriber. FAU provides a convenient way around this problem. By adding “.ezproxy.fau.edu” after “.com” or “.net” and before the rest of the URL I can access the resources through my university’s domain (once authenticated). This usually results in a way to download the publication if my university subscribes (which they often do).

When all else fails…

For newer publications that haven’t been indexed by other services yet Google is a good place to start. Simply searching for the title of a publication (of course, within quotes) is sometimes all that’s needed to find a a digital version. If this is not specific enough try including the first author’s last name.

My first journal publication, “An attention-driven model for grouping similar images with image retrieval applications”, has been accepted for publication in a special issue on Image Perception of the EURASIP Journal on Applied Signal Processing. It will appear in final form by the end of the year. In the meantime you can view the manuscript.

The paper presents a new method for extracting regions-of-interest from images. The high-level motivation is that users often seek objects within images and these objects are usually salient. The Itti-Koch model of visual attention plays a key role in the system. The resulting salient regions-of-interest can be clustered together for browsing or further computation.

The process of publishing a journal is far more arduous than for conference papers. Firstly, it takes a long time. This paper was originally submitted December 1st, 2005 and will appear nearly a year later. Secondly, the review process is far more thorough. The feedback we received from the reviewers was extremely value and contributed to a greatly improved revision of the paper.

I hope that this first journal on my dissertation work will be the first of several. I’m grateful to be part of a productive and highly-motivated research group. Research has many high and low points, but it’s extremely satisfying to be given confirmation that my work is headed in the right direction.

More information about my research is available online. My dissertation advisor is Dr. Oge Marques.