Video Compression - Since 1999

Most of the research was done at the Video and Image Processing Lab of the the University of California at Berkeley. The objective was to invent a new technique for the design of optimal two-dimensional functions used in Matching Pursuit based video codecs. These functions are optimal in the sense that they maximize the PSNR for a given bitrate.

Before 1999, Philippe Schmid-Saugeon has cross-checked several core experiments in the framework of the MPEG-4 standard definition.

Medical Devices - Since 1996

From 1996 to 1999, research was done at the Signal Processing Institute of the the Swiss Federal Institute of Technology at Lausanne, Switzerland. At that time most efforts were put on the design of a computer aided diagnosis system for skin cancer. Dermatoscopic (or dermoscopic) images were processed to mask out the lesions and automatically estimate relevant features. Algorithms had to be designed for segmentation, nonlinear filtering (mainly nonlinear isotropic or anisotropic filters), pattern analysis, symmetry analysis and quantification, etc. The successful results of this research opened the door to new opportunities in the industry.

The first of these opportunities was the design of a computer aided diagnosis system for cervical cancer. This research was done at Medispectra, Inc, and its objective was to detect cervical cancer from the time analysis of acetowhitening. A video sequence showing a cervix after application of acetic acid was acquired to this end. Image and signal processing algorithms were specifically designed to (1) align the images (tracking), (2) segment the images, (3) calibrate the images (real color calibration, not normalization), (4) measure the color of regions as a function of time, (5) extract relevant features from these functions. Based on these features, regions for which histopathology reports existed could be classified. The design of optimal classifiers was also a topic of this research. Early results were very encouraging.

The second opportunity was to design an in vivo system for the estimation of hematocrit. Two techniques were investigated: (1) the measure and analysis of optical density and (2) the use of confocal microscopy. In the second case most efforts were put into the design of a Monte Carlo simulation scheme to study light transport in skin with a confocal source-receptor configuration. This research was done for Intelligent Medical Devices, LLC.

In November 2005, Philippe Schmid-Saugeon joined Berata AG, a technology consulting firm based in Basel, Switzerland. Among other activities, he is involved in a major research project with Haag-Streit AG. He acts as an adviser for image processing technologies and directly participates to the design of image processing algorithms to process and analyze images produced with slit lamps. Major efforts are put on the evaluation of image sensors (CCD and CMOS) under poor lighting conditions (< 50 lux). Algorithms have already been designed for restoration (deblurring and denoising), segmentation, blending, etc.