Research

Dr. Hayes began his career in research in the general field of digital signal processing and, more specifically, in image and video processing. His early research activities have been in the following areas:

Signal Reconstruction.
Beginning with his doctoral research at M.I.T. on the reconstruction of multidimentsional signals from partial information (Fourier transform phase or magnitude), Dr. Hayes has published extensively in the area of signal reconstruction, constrained iterative deconvolution, and iiterative algorithms for signal enhancement.
Image and Video Compression
Dr. Hayes has worked on a variety of different problems related to image and video compression, including stereo image and video compression, multiview image compression, fractal compression of speech, images, and video. He has developed new compression algorithms using adaptive sampling and classified vector quantization, and worked on multiiframe pel-recursive algorithms for displacement estimation, and segmentation-based coding of motion difference and motion field images.
Face Detection and Face Recognition
Face detection and face recognition are important and long-standing research problems that Dr. Hayes and his students have made significant contributions to. To address the many difficult problems associated with robust face recognition in varying illuminations and poses, he and his students have developed a number of novel approaches to the problem including the use of Hidden Markov Models (HMMs), segmented linear subspaces, adaptive active appearance models, and eigenface-based super-resolution.
Video Analysis and Multimedia Signal Processing
Dr. Hayes has worked on number of research projects related to video analysis and multimedia signal processing. These projects include problems in scene change detection in video, scene segmentation for content-based indexing in the compressed domain, video object segmentation in the compressed domain, and content-based indexing and retrieval.
Adaptive Filtering and Adaptive Signal Processing
Dr. Hayes has a strong interest in adaptive signal processing and the use and application of adaptive filters. He has worked on signal processing problems that involve nonstational signals such as modeling of nonstationary random signals, spectral line tracking, iterative harmonic decomposition of nonstationary processes, and the use of Kalman filters for estimation and tracking,
Linear and Nonlinear Signal Modeling
While many of the research projects undertaken by Dr. Hayes have involved modeling of one-dimensional and multi-dimensional signals, some projects have focussed solely on signal modeling. These include ARMA modeling of time-varying signals using lattice filters, fractal modeling of speech, images, and video, and multipulse LPC coding.

While teaching and doing research in Seoul, Korea from 2011-2014, he became involved in projects related to image processing on digital cameras. These projects included depth estimation using dual off-axis color filter aperture cameras, spatially adaptive contrast enhancement, image defogging, and moving object detection using unstable cameras for video surveillance.

More recently, Dr. Hayes has become interested in the problem of Learning From Data, which is the title of a graduate level course that he has taught in Korea and is currently teaching at Geoge Mason University. This area is more often referred to as machine learning, and the theory and algorithms that have been developed over the past two decades have found applicationsin just about every corner of science and technology, finance to security and from communications to circuit design. He has developed a fully asynchronous course consisting of nearly 100 video modules and is preparing a textbook based on this course.

Monson H. Hayes

Research

Related links