Muhammed Ibrahim Sezan

Detection of slow-motion replay segments in sports video for highlights generation

We present a novel method for generating sports video summary highlights. Specifically, our method localizes semantically important events in sport programs by detecting slow motion replays of these events, and then generates highlights of these events at multiple levels. In our method, a hidden Markov model (HMM) is used to model slow motion replays, and an inference algorithm is introduced which computes the probability of a slow motion replay segment, and localizes the boundaries of the segment as well. An effective new feature is used in our HMM, based on a moving measure of the number of zero-crossings and the amplitudes of variations over time of video field differences. Furthermore, the method is capable of filtering out slow motion play segments in commercials. As compared with existing methods for video event detection, our method is more generic (ie, domain independent), and has the ability to capture inherently important events.

Image retrieval using blob histograms

We present a new method for image indexing and retrieval that is based on pixel statistics from varying spatial scales. The proposed method employs a structuring element to determine the frequency distribution of pixels locally in the image and to detect local groups of pixels with uniform color or texture attributes. The frequency distribution and relative sizes of such groups are summarized into a table termed as a blob histogram. By embedding spatial information, color blob histograms are able to distinguish images that have the same color pixel distribution but contain objects with different sizes or shapes, without the need for segmentation. Using isotropic structuring elements, blob histograms are invariant to rotations and translations of the objects in an image. Experimental results of using blob histograms in image retrieval are given in the paper.

Multimedia content recommendation engine with automatic inference of user preferences

We propose novel algorithms for automatically determining a users profile from his/her content usage history (profiling agent), and automatically filtering content according to the users profile (filtering agent). A fuzzy inference system is used to construct and periodically update the preferences of a user based on the users interactions with various types of content over an observation period. The proposed algorithms are designed to support an MPEG-7 or TV-anytime-compliant description framework, although they can also be utilized in any non-standard environment that provides structured descriptions of multimedia content.