Dive Deep / Learn New Skills

EI Short Courses complement the technical papers and social programs with a series of 2- and 4-hour classes taught by leading academic and industry experts from across the globe. Courses move from introductory to advanced in content, as well as from theoretical to practical, allowing attendees the opportunity to explore new areas or gain more depth in others.

In addition to offering the 15 in-person courses detailed below, this year EI is experiment with offering two online options with and in-person component. In this case, students may take the class from a location of their choice or join other participants in a room at the conference venue to experience the online class together. While the instructor will be presenting the class remotely, there will be an in class monitor for those participating onsite.

COURSES BY DAY / TIME

Please note: Course descriptions are found below the grid, and link from it. We are offering two HYBRID courses. For these, you may attend online from anywhere in real time or join others at the conference venue to network and watch the course together.

Sunday 21 January
8:00-10:15	10:30-12:15	13:30-15:30	15:45-17:45
IMAGE QUALITY
SC01: Image Quality Foundations: Standards for Mobile, Automotive, and Machine Vision Applications Peter Burns and Don Williams		NEW SC06: Hands-on Practice with Automotive Image Quality Measurements Laurent Chanas
IMAGE PROCESSING
SC02: Perceptual Metrics for Image and Video Quality: From Perceptual Transparency to Structural Equivalence Thrasyvoulos N. Pappas and Sheila S. Hemami		SC07 (HYBRID): CANCELED Advanced Image Enhancement and Deblurring Majid Rabbani
IMAGING		IMAGE QUALITY
SC03 (HYBRID): The Photon Path: Color Imaging From End to End Charles Poynton and David LeHoty		NEW SC08: Measurement and Calibration of Image Quality Key Performance Indicators (KPIs) Kevin J. Matherson and Uwe Artmann
ARTIFICIAL INTELLIGENCE (AI)
SC04: Fundamental Building Blocks of CNNs and LLMs Ray Ptucha and Clark Hochgraf		NEW SC09: Multimedia Forensics: From Forgery to the Deepfake Phenomenon Sebastiano Battiato and Luca Guarnera
MACHINE LEARNING
	SC05: CANCELED Introduction to Blockchain Gaurav Sharma	SC10: CANCELED Introduction to Probabilistic Models for Machine Learning Gaurav Sharma

Monday 22 January		Tuesday 23 January
8:30-10:30	10:45-12:45	8:30-10:30	10:45-12:45
IMAGING		COMPUTER VISION
SC11: Production Line Camera Color Calibration Eric Walowit and Dietmar Wueller		NEW SC14: Computer Vision and Analysis of Art David Stork
IMAGE PROCESSING		IMAGING
NEW SC12: Low Light Image Signal Processing (ISP) Stanley Chan		SC15: 3D Imaging Systems Hardware and Its Calibration Kevin J. Matherson
ARTIFICIAL INTELLIGENCE (AI)		IMAGING
NEW SC13: CANCELEDTowards Robust Neural Networks: Explainability, Uncertainty, and Intervenability Ghassan AlRegib and Mohit Prabhushankar		SC16: High-Dynamic-Range Theory and Technology Alessandro Rizzi and John McCann

Wednesday 24 January

8:30-10:30

IMAGE QUALITY

SC17:
Camera Noise Sources and Their Characterization Using International Standards
Kevin J. Matherson and Uwe Artmann

EI Short Course Descriptions WORKING AREA

Sunday 21 January

8:00 - 12:15 (4 hours)
Instructors: Peter D. Burns, Burns Digital Imaging, and Don Williams, Image Science Associates

SC01: Image Quality Foundations: Standards for Mobile, Automotive, and Machine Vision Applications [+/–]

Level: Introductory

Prerequisites: An introduction to methods for imaging performance testing (optical distortion, color-error, MTF, etc.) will be useful, but not required.

Benefits This course enables the attendee to:

Understand current methods for objective image quality evaluation.
Describe motivation for measuring imaging performance ‘by the numbers.’
Interpret current methods for objective image quality evaluation.
Critique the design of test targets used for ground-truth performance.
Describe why standard performance methods might differ with markets.
Identify challenges, and approaches, e.g., for evaluating wide Field-of-View (FOV) cameras.

The course starts with a discussion of objective image quality methods, as developed for image capture systems. Several of these methods have been adapted in emerging standards for, e.g., automotive (ADAS) and machine-vision applications. The instructors describe how and why imaging performance methods are being adopted. Most efforts rely on several ISO-defined methods, e.g., for color-encoding, image resolution, distortion, and noise. While several measurement protocols are similar, the image quality needs are different. For example, the EMVA 12288 standard for machine vision emphasizes the detector signal, while IEEE/P2020 includes optical, environmental, and motion conditions. This course is intended to provide a background for the intermediate course, SC02 Hands-on Image Quality Measurements for Computer Vision.

Intended Audience: Image scientists, quality engineers, and others evaluating digital camera and scanner performance.

Peter Burns is a consultant for imaging system evaluation, modeling, and design. Previously he worked for Carestream Health, Xerox, and Eastman Kodak. A frequent speaker at technical conferences, he has taught imaging courses for clients and universities for many years. He studied electrical engineering at Clarkson University, and completed his PhD in imaging science at RIT.

Don Williams is the founder of Image Science Associates, which focuses on quantitative imaging performance/fidelity evaluation for digital capture systems. His clients include national libraries, museums, and those with dental, mobile, and automotive applications. He contributes to several international standards activities. Williams studied imaging science at RIT, and previously worked at Eastman Kodak Research Labs.

8:00 - 12:15 (4 hours)
Instructors: Thrasyvoulos N. Pappas, Electrical and Computer Engineering, Northwestern University, and Sheila S. Hemami, independent consultant

SC02: Perceptual Metrics for Image and Video Quality: From Perceptual Transparency to Structural Equivalence [+/–]

Level: Intermediate

Prerequisites: Background in signal processing and basic statistics: frequency-based representations, filtering, distributions. Basic understanding of image compression.

Benefits This course enables the attendee to:

Gain a basic understanding of the properties of the human visual system and how current applications (image and video compression, restoration, retrieval, etc.) attempt to exploit these properties.
Gain an operational understanding of existing perceptually-based and structural similarity metrics, the types of images/artifacts on which they work, and their failure modes.
Understand current distortion models for different applications, and how they can be used to modify or develop new metrics for specific contexts.
Understand the differences between sub-threshold and supra-threshold artifacts, the HVS responses to these two paradigms, and the differences in measuring such responses.
Understand criteria by which to select and interpret a particular metric for a particular application.
Understand the capabilities and limitations of full-reference, limited-reference, and no-reference metrics, and why each might be used in a particular application.

Course Description
This course examines objective criteria for image quality based on models of visual perception. The primary emphasis is on fidelity but the instructors broaden the scope to include structural equivalence. They also discuss no-reference and limited-reference metrics and examine applications with emphasis on image and video compression. Also examined are near-threshold perceptual metrics, which estimate just-noticeable distortion thresholds, and supra-threshold metrics. The instructors consider metrics for structural equivalence, whereby the original and the distorted image have visible differences but both are of high visual quality. Also considered are procedures for evaluating the performance of quality metrics, including database design, models for generating realistic distortions, and subjective procedures for metric development and testing. The state-of-the-art and directions for future research are discussed.

Intended Audience: Image and video compression specialists who wish to gain an understanding of how performance can be quantified; engineers and scientists who wish to learn about objective image and video quality evaluation, managers who wish to gain a solid overview of image and video quality evaluation, students who wish to pursue a career in digital image processing, intellectual property and patent attorneys who wish to gain a more fundamental understanding of quality metrics and the underlying technologies; government laboratory personnel who work in imaging.

Thrasyvoulos N. Pappas received his SB, SM, and PhD in electrical engineering and computer science from MIT. From 1987 until 1999, he was a member of the technical staff at Bell Laboratories; he is currently a professor in the Department of Electrical and Computer Engineering at Northwestern University. His research interests are in image and video quality and compression, image and video analysis, content-based retrieval, perceptual models for multimedia processing, model-based halftoning, and tactile and multimodal interfaces. Pappas has served as vice-president publications IEEE Signal Processing Society, editor-in-chief of the IEEE Transactions on Image Processing , elected member of the Board of Governors of the Signal Processing Society of IEEE, chair of the IEEE Image and Multidimensional Signal Processing (now IVMSP) Technical Committee, technical program co-chair of ICIP-01 and ICIP-09, and co-chair of the 2011 IEEE IVMSP Workshop on Perception and Visual Analysis. He has also served as co-chair of the SPIE/IS&T Electronic Imaging Symposium and co-chair of the EI Conference on Human Vision and Electronic Imaging. He is currently co-editor-in-chief of the IS&T Journal of Perceptual Imaging. Pappas is a Fellow of IEEE, SPIE, and IS&T.

Sheila Hemami holds a BSEE from the University of Michigan, and an MSEE and PhD from Stanford University. She worked at Hewlett-Packard Laboratories and was a professor and associate director of the School of Electrical and Computer Engineering at Cornell University and professor and chair of the Department of Electrical & Computer Engineering at Northeastern University. More recently, she was director of strategic technical opportunities at the Charles Stark Draper Laboratory, where she created and led a new business unit dedicated to large-scale challenges for the environment and for under-served medical populations. In 2021, she joined Triple Ring Technologies to lead the company’s Boston office. Hemami's research interests broadly concern communication of visual information from the perspectives of both signal processing and psychophysics. She has held various visiting positions, most recently at INRIA Bretagne in Rennes, France. She was elected a Fellow of the IEEE in 2009 for her contributions to robust and perceptual image and video communications. She has received numerous university and national teaching awards, including Eta Kappa Nu's C. Holmes MacDonald Award. She served as an IEEE vice-president publications products and services, was a distinguished lecturer for the IEEE Signal Processing Society, and editor-in-chief for the IEEE Transactions on Multimedia. She has held various technical leadership positions in the IEEE.

8:00 - 12:15 (4 hours)
Instructors: Charles Poynton, independent researcher, and David LeHoty, independent consultant

SC03: The Photon Path: Color Imaging from End to End [+/–]

Level: Advanced

Prerequisites: Intermediate knowledge of physics, math, and color imaging science.

Benefits This course enables the attendee to:

Understand the imaging path end-to-end: from sensors, through signal processing, to displays. Understand color transforms and the fundamentals of color appearance models.
Understand the imaging pipeline, including OETFs, EOTFs, and their inverses.

Course Description
In 2020, The Technical Committee of IMAGO (the International Federation of Cinematographers) undertook creation of a glossary to clarify digital imaging technologies, along with a photon path diagram to tie together the various key concepts. The work is now published, and has application well beyond cinematography.

This online course uses the glossary and the photon path diagram to explore the deep principles of imaging systems, from sensors, through signal processing, to displays. The instructors discuss fundamental color science – trichromaticity, metamerism, and the necessity for signal processing at or near the camera to create signals referenced to physical additive primaries. They also cover topics including sensor spectral response and the inevitable camera metamerism; classic colorimetry and the computation of 3×3 matrices, both the color formation matrix (CFM) associated with cameras and the transforms among color spaces in processing and at display; and the various color spaces and the transforms among them. The course discusses color appearance models, and the important issue of picture rendering (whereby color signals are altered to compensate for differences between scene conditions and display conditions), as well as the emergent technologies of high dynamic range (HDR) and wide color gamut (WCG).

Intended Audience: Color image scientists, engineers, programmers, and technologists.

Charles Poynton is an independent image and color scientist specializing in the physics, mathematics, and engineering of digital color imaging systems, including HD/HDTV/UHD/4K/8K, and D-cinema, high dynamic range (HDR), and wide color gamut (WCG) systems. He does technology forecasting, systems modeling, algorithm development, video signal processing architecture, and color characterization and calibration. He is the author of the book Digital Video and HD Algorithms and Interfaces. He has a PhD from Simon Fraser University.

David LeHoty is an independent consultant in Mountain View, California. His areas of expertise include D-cinema and phone displays, color measurements, calibration, LED balancing, gamut coverage, luminance non-uniformity, color space conversion, video signal processing, color difference errors, and manufactured product color error trends. LeHoty earned a master’s degree in electrical engineering from Stanford University and holds 15 US patents.

8:00 - 12:15 (4 hours)
Instructors: Ray Ptucha, Apple, Inc., and Clark Hochgraf, Rochester Institute of Technology

SC04: Fundamental Building Blocks of CNNs and LLMs [+/–]

Level: Intermediate

Prerequisites: Prior familiarity with basic machine learning and a scripting language are helpful.

Benefits This course enables the attendee to:

Become familiar with deep learning concepts and applications.
Understand how deep learning methods, specifically convolutional neural networks and large language models (LLMs), work.
Learn how to build, test, and improve the performance of deep networks using popular open- source utilities.

Course Description
We are fortunate to witness how deep learning is forever changing society and our lives. These machine learning models are auto magical, but they are not magic. This tutorial investigates two of the most popular paradigms: convolutional neural networks (CNNs) and recurrent large language models (LLMs) and looks at key building blocks that support these models.

Modern CNNs can be just as good as visual transformers for object detection and segmentation. This course explains what makes these CNN models tick, and helps students appreciate where the community has been and where it is headed.

Large language models and transformers, such as those used in ChatGPT, have revolutionized the sequential and static recognition communities. The course starts with sequential models, introduces attention and transformers, and then discusses how and why LLMs are so powerful. Through this course, students discover the structure of LLMs and their training methods, strengths, and weaknesses and also learn how to leverage APIs to harness the power of generative AI.

Intended Audience: Engineers, scientists, students, and managers interested in acquiring a broad understanding of deep learning.

Ray Ptucha is a computational display technology leader in the Visual Experience Group at Apple, where he is responsible for machine learning and algorithms in display products. He was an associate professor in computer engineering and director of the Machine Intelligence Laboratory at Rochester Institute of Technology (RIT), where he co-authored more than 100 publications including topics in machine learning, computer vision, and robotics, with a specialization in deep learning. Prior to RIT, Ptucha was a research scientist with Eastman Kodak Company, where he worked on computational imaging algorithms and was awarded 40 US patents. He graduated from SUNY/Buffalo with a BS in computer science and a BS in electrical engineering. He has earned an MS in image science and a PhD in computer science from RIT. Ptucha was awarded an NSF Graduate Research Fellowship in 2010, and his PhD research earned the 2014 Best RIT Doctoral Dissertation Award. He is a passionate supporter of STEM education and has served as an NVIDIA certified Deep Learning Institute instructor and chair of the Rochester area IEEE Signal Processing Society. He is an active supporter of his local IEEE chapter and FIRST robotics organizations.

Clark Hochgraf is an associate professor with the Department of Electrical, Computer, and Telecommunications Engineering Technology at the Rochester Institute of Technology. He received a PhD in electrical engineering from the University of Wisconsin-Madison. His areas of research interest include intelligent vehicles, power conversion, and education of engineers.

10:30 - 12:15 (2 hours)
Instructor: Gaurav Sharma, University of Rochester

CANCELED SC05: Introduction to Blockchain [+/–]

Level: Introductory

Benefits This course enables the attendee to:

Distinguish between centralized and distributed ledgers and highlight their pros and cons.
Describe proof-of-work and proof-of-stake approaches for consensus and the pros and cons for each of these.
Explain the utility and applicability of blockchains in diverse applications.
Cite example applications of blockchains.

Course Description
This course introduces attendees to blockchains, which have emerged as a revolutionary technology that has the potential to disrupt a range of diverse business processes and applications. Using a concrete application setting, the course illustrates the construction of blockchains as a distributed, secure, and tamper-resistant framework for the management of transaction ledgers. Necessary background concepts in the technologies underlying blockchains, including basic cryptographic tools, are introduced as required. Current and emerging applications of blockchains are surveyed, including those in media security.

Intended Audience: Engineers, scientists, students, and managers interested in understanding how blockchains are constructed and how they can be useful in a variety of business processes. The course includes an overview of necessary background information, including cryptographic tools utilized in the blockchain; prior familiarity with these concepts is not required.

Gaurav Sharma is a professor of electrical and computer engineering and of computer science at the University of Rochester, where his research spans signal and image processing, computer vision, color imaging, media security, and bioinformatics. He has extensive experience in media security across diverse application domains. Prior to joining the University of Rochester, he was a principal scientist and project leader at the Xerox Innovation Group. Additionally, he has consulted for several companies on the development of image processing and computer vision algorithms. He holds 55 issued patents and has authored over 220 peer-reviewed publications. He is editor of the Digital Color Imaging Handbook published by CRC Press. He was the editor-in-chief for the IEEE Transactions on Image Processing from 2018-2020, and for the IS&T/SPIE Journal of Electronic Imaging from 2011 through 2015. Sharma is a fellow of IS&T, IEEE, and SPIE.

13:30 - 17:45 (4 hours)
Instructor: Laurent Chanas, DXOMARK Image Labs

SC06: Hands-on Practice with Automotive Image Quality Measurements [+/–]

Level: Intermediate (beginners accepted)

Prerequisites: Fundamentals of image processing and camera inner workings.

Benefits This course enables the attendee to:

Identify the most important image quality attributes for computer vision.
Understand the problems encountered when testing a real device in a lab and how to approach them.
Understand which influencing factor one needs to take care of for each measurement
See a real-life measurement setup and how to control it.
Discuss with experts who have participated in the standardization of the metrics and the development of the equipment.

Course Description
The purpose of this course is to give a short overview of what to measure and how to measure it in terms of image quality for computer vision. An automotive application is the case study of focus. Several aspects of typical automotive systems make them very challenging to measure, more so than the consumer imaging systems to which most people are accustomed. In fact, when designing such a measurement, one is quickly confronted with two main aspects: A much wider dynamic range than most consumer imaging systems and a much larger field of view.

This course describes how the definition and measurement setup and protocols of even well-known metrics (such as noise) needed to be extended to cope with these new factors in the new IEEE/P2020 standard. Half the the course is dedicated to the practical challenges and solutions involved in making an accurate measurement in a laboratory. The IEEE/P2020 setups are present in the room. The aim is to give attendees an overview of all the influencing factors to take into account for each measurement.

Intended Audience: Image scientists, engineers, or managers who wish to learn more about image quality for computer vision.

Laurent Chanas has been working in image quality assessment and improvement for more than 20 years. He is now responsible for the image quality assessment solution, Analyzer. Previously, Chanas worked at DxO Labs, developing image processing algorithms (including several international patents) and managing the test and measurement laboratory. Chanas has a master’s from the Ecole Nationale de l’Electronique et ses Applications, as well as a PhD in image processing from University Cergy-Paris.

13:30 - 17:45 (4 hours)
Instructor: Majid Rabbani, Rochester Institute of Technology

CANCELED SC07: Advanced Image Enhancement and Deblurring [+/–]

Level: Advanced

Prerequisites: Knowledge of basic image processing and linear systems (e.g., convolution, Fourier Transforms, etc.) would be very useful.

Benefits This course enables the attendee to:

Understand advanced algorithms used for contrast enhancement such as CLAHE, Photoshop Shadows/Highlights, and Dynamic Range Compression (DRC).
Understand advanced techniques used in image sharpening such as variations of nonlinear unsharp masking.
Understand advanced techniques used in image noise removal such as bilateral filtering and CNN-based techniques.
Understand how motion information can be utilized in image sequences to improve the performance of various enhancement techniques such as sharpening, noise removal, and artifact removal.
Understand Wiener filtering and its variations for image deblurring (restoration).

Course Description
This online course discusses some of the advanced algorithms used for contrast enhancement, noise reduction, sharpening, and deblurring of still images and video. Applications include consumer and professional imaging, medical imaging, forensic imaging, surveillance, and astronomical imaging. Image examples complement the technical descriptions.

Intended Audience: Scientists, engineers, and technical managers who need to understand and/or apply the techniques employed in digital image processing in various products in a diverse set of applications such as medical imaging, professional and consumer imaging, and forensic imaging will benefit from this course. Some knowledge of digital filtering (convolution) and frequency decomposition is necessary for understanding the deblurring concepts.

Majid Rabbani has 38 years of experience in digital imaging. After a 33-year career at Kodak Research labs, he retired in 2016 with the rank of Kodak Fellow. Currently, he is a professor of practice in the Electrical and Microelectronic Department at Rochester Institute of Technology (RIT). He is a twice co-recipient of the Kodak C. E. K. Mees Research Award and the co-recipient of two Engineering Emmy Awards. He is a Fellow of IEEE and SPIE, a Kodak Distinguished Inventor with 44 issued patents, and the 2015 recipient of the Electronic Imaging Scientist of the Year Award. He has been an active educator in the digital imaging community for the past 39 years.

13:30 - 17:45 (4 hours)
Instructors: Kevin J. Matherson, Microsoft, and Uwe Artmann, Image Engineering GmbH & Co KG

SC08: Measurement and Calibration of Image Quality Key Performance Indicators (KPIs) [+/–]

Level: Intermediate

Prerequisites: No specific imaging background is assumed. It will be helpful if participants have a basic knowledge of engineering and physics.

Benefits This course enables the attendee to:

Understand image quality parameters that should be tested or calibrated during the development/production of a camera
Be able to explain and demonstrate a basic understanding of the required measurement and calibration algorithms.
Be able to read, interpret, and understand image-quality-related specifications for a given supply chain.
Perform the various calibrations typically done in camera modules (intrinsic calibration, distortion calibration, relative illumination, color shading, spectral calibrations, gain, actuator variability, etc.)

Course Description
Every camera produced needs some form of image quality check and calibration. This course discusses and explains the different image quality parameters, hardware calibrations, their purpose, and how they are measured.

This course discusses camera parameters that typically vary enough to require individual calibration or are checked for pass/fail criteria. Also discussed are those that should be class calibrated or can be corrected using the image pipeline. As different manufacturers in the camera component supply chain have different specifications and use different terminology, this course helps students to understand camera sub-components and how they relate to overall camera requirements much better, as well as to obtain a good understanding of whether to test or calibrate a given component.

Intended Audience: Everyone who has a technical background and is new to the imaging industry. Also, program managers, project leads, or engineers working on the camera system benefit from the overview of different aspects of camera measurement and calibration.

Kevin Matherson is a director of optical engineering at Microsoft Corporation working on advanced optical technologies for AR/VR, machine vision, and consumer products. Prior to Microsoft, he participated in the design and development of compact cameras at HP and has more than 20 years of experience developing miniature cameras for consumer products. His primary research interests focus on sensor characterization, optical system design and analysis, and the optimization of camera image quality. Matherson holds an MS and PhD in optical sciences from the University of Arizona.

Uwe Artmann studied photo technology at the University of Applied Sciences in Cologne following an apprenticeship as a photographer, and finished with the German ’Diploma Engineer’ degree. He is now CTO at Image Engineering, an independent test lab for imaging devices and manufacturers of all kinds of test equipment for these devices. He is an active contributor to different international standards committees including IEEE-P2020, EMVA1288, and VCX Forum.

13:30 - 17:45 (4 hours)
Instructors: Sebastiano Battiato and Luca Guarnera, University of Catania, Department of Mathematics and Computer Science

SC09: Multimedia Forensics: From Forgery to the Deepfake Phenomenon [+/–]

Level: Intermediate

Prerequisites: Basic knowledge of signal processing.

Benefits This course enables the attendee to:

Acquire basic knowledge of multimedia forensics.
Learn those main technologies behind multimedia content manipulations, starting from classic image/video forgery attacks to modern deepfakes creation algorithms.
Be able to analyze multimedia content in detail in order to identify those intrinsic and discriminative manipulation traces found in the data under analysis and invisible to the human eye.

Course Description
This course covers the fundamental aspects of multimedia forensics, image/video forgery, and deepfakes creation and detection. Course participants acquire a foundational understanding of these critical subjects. Participants explore image and video forgery techniques, gaining insights into how these manipulations occur and how to detect them effectively. A significant portion of the course is dedicated to generative models and the deepfake phenomenon, elucidating their role in creating convincing synthetic media. The course delves into the primary approaches for deepfake detection, enabling students to recognize and combat the proliferation of fraudulent content. Ultimately, this course equips individuals with the knowledge and tools needed to counter the illicit use of these powerful technologies.

Intended Audience: PhD students, researchers.

Sebastiano Battiato is a full professor, scientific coordinator of the PhD program in computer science, and deputy director for strategic planning and information systems at the University of Catania, Italy. His research interests include computer vision, imaging technology, and multimedia forensics. He is involved in research and directorship of the IPLab research lab. An IEEE senior member, he has edited six books, co-authored about 300 papers, is a co-inventor on 22 international patents, and has chaired several international events. Battiato was director of the International Computer Vision Summer School (ICVSS) for many years. Battiato has been involved in the evaluation process of research projects on behalf of national and international institutions and has worked as a consultant on behalf of the public prosecutors of many Italian regions on issues related to imaging in the forensic field. Since 2016 he has been the founder and scientific advisor of iCTLab (www.ictlab.srl) – a spin-off of the University of Catania that operates in the field of digital forensics.

Luca Guarnera is a research fellow in computer science at the University of Catania. He received a PhD in computer science from the Department of Mathematics and Computer Science, University of Catania. Part of the PhD research was carried out at the University of Hertfordshire, College Lane Campus, Hatfield, UK, under the supervision of Professor Salvatore Livatino, working on the creation of software for forensic ballistics analysis and firearms comparison, using the Oculus Rift S headset. He received his MSc in computer science from the Department of Mathematics and Computer Science, University of Catania. He joined IPLab in 2015. In 2017 and 2019, he took part in two Mohamed Bin Zayed International Robotics Challenge (MBZIRC) international robotics competitions. He also participated in four editions of the International Computer Vision Summer School (ICVSS), one edition of Medical Imaging Summer School (MISS), and one edition of summer school on signal processing (S3P). His main research interests are computer vision, machine learning, and multimedia forensics and its related fields with a focus on the deepfake phenomenon.

13:30 - 17:45 (4 hours)
Instructor: Gaurav Sharma, University of Rochester

CANCELED SC10: Introduction to Probabilistic Models for Machine Learning [+/–]

Level: Introductory

Prerequisites: Familiarity with basic probability concepts.

Benefits This course enables the attendee to:

Describe and intuitively explain fundamental probabilistic concepts such as independence, Bayes' rule, and stationarity.
Explain the basis of Maximum Aposteriori Probability (MAP) and Maximum Likelihood (ML) detection and estimation rules.
Describe how latent variables and sequential dependence underlie expectation maximization and hidden Markov Models.
Develop simple applications of probabilistic models for computer vision and image processing problems.
Cite and explain application examples involving the use of probabilistic models in computer vision, machine learning, and image processing.

Course Description
This course provides attendees with a foundation in inference and estimation for machine learning using probabilistic models. Starting from the broad base of probabilistic inference and estimation, the course develops the treatment of specific techniques that underlie many current day machine learning and inference algorithms. Topics covered include a review of concepts from probability and stochastic processes, IID and Markov processes, basics of inference and estimation, Maximum Aposteriori Probability (MAP) and Maximum Likelihood (ML), expectation maximization for ML estimation, hidden Markov Models, and Markov and conditional random fields. The pedagogical approach is to illustrate the use of models via concrete examples: Each model is introduced via a detailed toy example and then illustrated via one or two real-world application examples. The presentation of the core material is cohesive and self-contained with an overview and pointers to resources for extending knowledge beyond the course.

Intended Audience: Engineers, scientists, students, and managers interested in understanding how probabilistic models are used in inference and parameter estimation problems in today's machine learning and computer vision applications and in applying such models to their own problems. Prior familiarity with basics of probability and with matrix vector operations are necessary for a thorough understanding, although a short overview is provided for attendees lacking this background, which should allow them to also develop an intuitive high-level understanding.

Gaurav Sharma is a professor of electrical and computer engineering and of computer science at the University of Rochester, where his research spans signal and image processing, computer vision, color imaging, media security, and bioinformatics. He has extensive experience in media security across diverse application domains. Prior to joining the University of Rochester, he was a principal scientist and project leader at the Xerox Innovation Group. Additionally, he has consulted for several companies on the development of image processing and computer vision algorithms. He holds 55 issued patents and has authored over 220 peer-reviewed publications. He is editor of the Digital Color Imaging Handbook published by CRC Press. He was the editor-in-chief for the IEEE Transactions on Image Processing from 2018-2020, and for the IS&T/SPIE Journal of Electronic Imaging from 2011 through 2015. Sharma is a fellow of IS&T, IEEE, and SPIE.

Monday 22 January

8:30 - 12:45 (4 hours)
Instructors: Eric Walowit, consultant, and Dietmar Wueller, Image Engineering GmbH & Co. KG

SC11: Production Line Camera Color Calibration [+/–]

Level: Intermediate

Benefits This course enables the attendee to:

Understand the need for camera colorimetric characterization and the impact of color calibration on image quality and manufacturing yield.
Perform target-based and spectral-based camera characterization.
Solve for colorimetric camera transforms and build profiles using linear and nonlinear techniques.
Evaluate current colorimetric camera characterization hardware and software technology and products.
Participate in hands-on spectral camera characterization, camera transform generation, and matching from capture to display.

Course Description
This course covers the process of colorimetric camera characterization in theory and practice. The need for camera characterization and calibration and the impact on general image quality are first reviewed. Known issues in traditional approaches are discussed. Methodology for building camera colorimetric transforms and profiles are detailed step-by-step. State-of-the-art solutions using current technology are presented including monochromators, multispectral LED light sources, in situ measurements of spectral radiances of natural objects, and modern color transform methods including multidimensional color lookup tables. A live demonstration is performed of the end-to-end process of spectral camera characterization, camera transform generation, and matching from capture to display. This course provides the basis needed to implement advanced color correction in cameras and software.

Intended Audience: Engineers, project leaders, and managers involved in camera image processing pipeline development, image quality engineering, and production-line quality assurance.

Eric Walowit’s interests are in color management, appearance estimation, and image processing pipelines for digital photographic applications. He is the founder (retired) of Color Savvy Systems, a color management hardware and software company. He graduated from RIT’s image science program, concentrating in color science. Walowit is a member of ICC, ISOTC42, IS&T, and CIEJTC10.

Dietmar Wueller studied photographic sciences in Cologne, Germany. He is the founder of Image Engineering, one of the leading suppliers for test equipment for digital image capture devices. Wueller is a member of IS&T, DGPH, and ECI and is the German representative for ISO/TC42/WG18. He also participates in several other standardization activities.

8:30 - 10:30 (2 hours)
Instructor: Stanley Chan, Purdue University
SC12: Low Light Image Signal Processing (ISP) [+/–]

Level: Introductory

Benefits This course enables the attendee to:

Understand the theoretical limits of imaging at low light.

Be up to date on the classical and deep learning techniques for image denoising.

Describe the noise models and their impacts to downstream ISP tasks.

Course Description
The continuous shrinkage of pixel pitch has led to a smaller full well capacity, causing the signal-to-noise to drop when imaging in low-light conditions. Overcoming the photon shot noise has been the biggest challenge in imaging because shot noise is present even if the sensors are perfect. This short course provides an overview of image denoising in the low-light regime. Topics to discuss include:
1. Image sensor model
2. Signal to noise ratio and dynamic range
3. Classical image denoising algorithms and their limit
4. Deep neural network denoiser
5. Low-light object detection and recognition techniques
6. Open discussions
Part of the materials is based on a new course the instructor teaches at Purdue University: https://engineering.purdue.edu/ChanGroup/ECE495.html
Many examples and worked problems will be discussed in this short course.

Intended Audience: Professionals working in the camera / image processing industry and graduate students interested in doing image processing and computational photography.

Stanley Chan is currently the Elmore Associate Professor of electrical and computer engineering at Purdue University. He received his PhD in electrical engineering and MA in mathematics from UC San Diego, and his BEng in electrical engineering from the University of Hong Kong. In 2012-2014, he was a postdoctoral research fellow at Harvard University. He is a senior member of IEEE. He is currently an Associate Editor of the IEEE Transactions on Computational Imaging, where he is recognized by the IEEE Signal Processing Society as an outstanding editorial board member. In IEEE, he was an elected member of the technical committee on computational imaging. He has also served as an associate editor for OSA Optics Express.

8:30 - 10:30 (2 hours)
Instructors: Ghassan AlRegib, Georgia Institute of Technology, and Mohit Prabhushankar, Georgia Institute of Technology
CANCELED SC13: Towards Robust Neural Networks: Explainability, Uncertainty, and Intervenability [+/–]

Level: Advanced

Prerequisites: A basic understanding of neural networks and image processing.

Benefits This course enables the attendee to:

Appreciate the necessity of robustness in neural networks when applied in real world scenarios like autonomous vehicles and chat bots

Concretely list the prerequisites for a robust autonomous system that uses neural networks

Implicitly compute contextual and relevant explanations for the decisions made by neural networks

Provide relevant uncertainties for the neural networks during decision making process

Imbue neural networks with the ability to be intervened on, like prompting and data deletion

Be introduced to relevant applications that deal with robustness including image quality assessment, out-of-distribution, adversarial, and anomaly detection

Course Description
Neural network driven applications like ChatGPT suffer from hallucinations where they confidently provide inaccurate information. A fundamental reason for this inaccuracy is the lack of robust measures that are applied on the underlying neural network predictions. This short course identifies and expounds on three interrelated robustness measures that every decision made by a neural network must be equipped with. Explainability and uncertainty research fields are accompanied by a large body of literature that analyzes decisions. Intervenability, on the other hand, has gained recent prominence due its inclusion in the GDPR regulations and a surge in prompting-based neural network architectures.
In this short course, the instructor connects all three fields using gradient-based techniques to create robust machine learning models.

Intended Audience: Graduate students, researchers, and engineers working in different topics related to visual information processing and analysis.

Ghassan AlRegib is currently the John and Marilu McCarty Chair Professor in the School of Electrical and Computer Engineering (ECE) at the Georgia Institute of Technology. In the Omni Lab for Intelligent Visual Engineering and Science (OLIVES), he and his group work on robust and interpretable machine learning algorithms, uncertainty and trust, and human in the loop algorithms. The group has demonstrated work on a wide range of applications such as autonomous systems, medical imaging, and subsurface imaging and is interested in advancing the fundamentals as well as the deployment of such systems in real-world scenarios. AlRegib has been issued several US patents and invention disclosures. He is a Fellow of IEEE and has served on the editorial board of several transactions and served as the TPC Chair for ICIP 2020, ICIP 2024, and GlobalSIP 2014. AlRegib was area editor for the IEEE Signal Processing Magazine. He received the 2008 ECE Outstanding Junior Faculty Member Award and the 2017 Denning Faculty Award for Global Engagement. He and his students received the ICIP 2019 Best Paper Award.

Mohit Prabhushankar received his PhD in electrical engineering from the Georgia Institute of Technology, where he is currently a postdoctoral research fellow in the Omni Lab for Intelligent Visual Engineering and Science (OLIVES). He works in the fields of image processing, machine learning, active learning, healthcare, and robust and explainable AI. Prabhushankar is the recipient of the ICIP 2019 Best Paper award and the ICIP 2020 Top Viewed Special Session Paper Award, as well as the ECE Outstanding Graduate Teaching Award, a CSIP Research award, and the Roger P. Webb ECE Graduate Research Assistant Excellence award, all in 2022.

Tuesday 23 January

8:30 - 12:45 (4 hours)
Instructor: David G. Stork, Stanford University

SC14: Computer Vision and Analysis of Art [+/–]

8:30 - 12:45 (4 hours)
Instructor: Kevin J. Matherson, Microsoft

SC15: 3D Imaging Systems Hardware and its Calibration [+/–]

Level: Intermediate

Prerequisites: Some imaging experience is helpful but not assumed.

Benefits This course enables the attendee to:

Describe fundamental principles of depth technology and 3D imaging.
Understand the trade-offs among currently available depth system technologies and determine which will best match a particular application.
Understand the key components of various depth technologies: optics, illuminators, sensors.
Understand concepts and design considerations for depth cameras: stereo, active stereo, structured light, time of flight.
Understand passive and active depth camera calibration.
Compare time-of-flight imaging to triangulation-based approaches.
Understand methods of benchmarking depth cameras.

Course Description
Image noise is one of the most important aspects of image quality. This course describes the sources of noise, what influences noise, and how one can measure it. Noise does have different sources in the fundamental physics of the imaging sensor and will be modified in different ways through the image pipeline from light in to digital value out. The course will help one understand the sources of noise and how this impacts imaging system design decisions. One will also learn how to properly measure noise in imaging systems. As there are different standards for noise measurement, each with its own specific setup, in which analysis and terminology differ, there can be confusion about what to measure and how. This course will help to clear up misunderstandings as well as provide hands-on experience in measuring noise.

Camera modules are now commonplace, integrated in devices ranging from mobile phones to automobiles. CMOS image sensor technology and advances in image processing technology, as well as advances in packaging and interconnect technology, have created the ability to use cameras in applications that were unheard of just a few years ago. Emerging applications of cameras include Internet of Things (IoT), biometrics, augmented reality, machine vision, medicine, and 3D imaging. There are a variety of different methods for three-dimensional imaging, each with their own strengths and weaknesses. This course provides an overview of the main methods for depth imaging: stereo, active stereo, structured light, and time of flight. Following a review of 2D camera fundamentals, the course compares the various architectures of depth cameras. Next, the course covers camera calibration, a crucial step in machine vision that is required for measurements of the environment. Finally, the course provides an overview of depth camera benchmarking, which adds more complexity to testing as compared to 2D cameras.

Kevin J. Matherson is a director of optical engineering at Microsoft Corporation working on advanced optical technologies for AR/VR, machine vision, and consumer products. Prior to Microsoft, he participated in the design and development of compact cameras at HP and has more than 20 years of experience developing miniature cameras for consumer products. His primary research interests focus on sensor characterization, optical system design and analysis, and the optimization of camera image quality. He is also an active participant in the standards community and the project leader for the proposed standard on depth camera metrology. Matherson holds an MS and PhD in optical sciences from the University of Arizona.

8:30 - 10:30 (2 hours)
Instructors: Alessandro Rizzi, University of Milano, and John J. McCann, McCann Imaging

SC16: High-Dynamic-Range (HDR) Theory and Technology [+/–]

Level: Overview

Benefits This course enables the attendee to:

Measure the optical limits in acquisition and display; in particular measure the scene-dependent effects of optical glare.
Compare the accuracy of scene capture using single and multiple exposures in normal and RAW formats.
Engage in a discussion of human spatial vision that responds to the retinal image altered by glare.
Engage in a discussion of current HDR TV systems and standards: tone-rendering vs. spatial HDR methods.
Explore the history of HDR imaging.

Course Description
To understand HDR imaging is to understand the properties and limitations of the artist’s goal, the calculated sensation goal, and the accurate scene radiance goal.

This course emphasizes measurements of physics (accurate reproduction) and psychophysics (visual appearance). Physics shows limits caused by optical glare; HDR does not reproduce scene radiances. Psychophysics shows that human vision’s spatial-image-processing renders scene appearance.

The course reviews successful HDR reproductions; limits of radiance reproduction; HDR TV’s technology and standards; appearance and display luminance; and appearance models. HDR technology is a complex problem controlled by optics, signal-processing, and visual limits. The solution depends on its goal: physical information or preferred appearance.

Intended Audience: Anyone interested in using HDR imaging: science, technology of displays, and applications, including students, color scientists, imaging researchers, medical imagers, software and hardware engineers, photographers, cinematographers, and production specialists.

Alessandro Rizzi is a full professor in the Computer Science Department of the University of Milano. Since 1990 he has pursued research in digital imaging with a particular interest in color, visualization, photography, and HDR. He is senior editor of Color Research and Applications and associate editor of the Journal of Electronic Imaging. He has been topical editor of the Journal of Optical Society of America, secretary of CIE Division 8, and and IS&T Vice President. He is an IS&T Fellow and in 2015 received the Davies Medal from the Royal Photographic Society.

John McCann worked in, and managed, Polaroid’s Vision Research Laboratory (1961-1996). He studied Retinex theory, color constancy, color from rod/cone interactions at low light levels, image reproduction, appearance with scattered light, cataracts, and HDR imaging. He is a Fellow of IS&T and the Optical Society of America (OSA); a past president of IS&T and the Artists Foundation, Boston; IS&T/OSA 2002 Edwin Land Medalist and IS&T 2005 Honorary Member. In 2021 he received the Judd Award at the AIC International Congress, Milan.

Wednesday 24 January

8:30 - 10:30 (2 hours)
Instructors: Kevin J. Matherson, Microsoft, and Uwe Artmann, Image Engineering GmbH & Co. KG

SC17: Camera Noise Sources and Their Characterization Using International Standards [+/–]

Level: Introductory

Prerequisites: No specific imaging background is assumed. It will be helpful if participants have a basic knowledge of engineering or physics.

Benefits This course enables the attendee to:

Explain the fundamental noise sources in imaging systems and what influences them.
Explain how various operations in the imaging pipeline can increase, decrease, or modify image noise.
Measure noise by applying the most important noise-related international standards.
Predict and model system-level camera performance using international standards.

Course Description
Image noise is one of the most important aspects of image quality. In this course, students learn the sources of noise, what influences noise, and how to measure it. Noise does have different sources in the fundamental physics of the imaging sensor and is modified in different ways through the image pipeline, from light in to digital value out. The course aids in the understanding of the sources of noise and how these sources impact imaging system design decisions. Students also learn how to properly measure noise in imaging systems. As there are different standards for noise measurement, each with its own specific setup, in which analysis and terminology differ, there can be confusion about what to measure and how. This course helps clear up misunderstandings and provides hands-on experience in measuring noise.

Intended Audience: Everyone involved in designing a camera system, from engineering level to project management.

Kevin Matherson is a director of optical engineering at Microsoft Corporation working on advanced optical technologies for AR/VR, machine vision, and consumer products. Prior to Microsoft, he participated in the design and development of compact cameras at HP and has more than 20 years of experience developing miniature cameras for consumer products. His primary research interests focus on sensor characterization, optical system design and analysis, and the optimization of camera image quality. Matherson holds an MS and PhD in optical sciences from the University of Arizona.

Uwe Artmann studied photo technology at the University of Applied Sciences in Cologne following an apprenticeship as a photographer, and finished with the German ’Diploma Engineer’ degree. He is now CTO at Image Engineering, an independent test lab for imaging devices and manufacturers of all kinds of test equipment for these devices. He is an active contributor to different international standards committees such as IEEE-P2020, EMVA1288, and VCX Forum.

Short Course Fees

	Through 31 December			Starting 1 January
	Member	Non-Member	Student	Member	Non-Member	Student
4-hour courses	$330	$355	$95	$380	$405	$120
4-hour HYBRID courses (SC03 and SC07 ONLY)	$255	$280	$75	$305	$330	$105
2-hour courses	$200	$225	$65	$250	$275	$90
Take 3 or more courses and receive 10% discount. Use Coupon Code 2024PICK3 at checkout.