The goal of color transfer is to transfer the color information from a source image to recolor or color a target image. Color transfer methods have great potential in many applications such as animation, video editing, image enhancement, photo editing software, color matching and image security. These different applications can be categorized into three groups: image coloring, image recoloring and image sequencing. For situations when the target image is a grayscale image, color transfer aims to utilize the color information from a color source image to provide color to the target image. When the target image is a color image, this color transfer aims to recolor it to make it capture and incorporate the color information from the source image. The last kind of application is to recolor the target image with a different combination of color information from more than one source image. This will output an image sequence showing the color gradually changing from the color information from one source image to another one of source images. We demonstrate our image sequencing is suitable for gradually changing video scenes from summer to winter and creating the in-between frames necessary to accomplish this.

• See related media ➟

What makes one picture of an object better than another? The resolution? The color? Dr. Panetta develops computer algorithms that enhance the visual quality of images so that they are best suited to human perception and allow the human eye to see things that wouldn't be visible in a raw photograph or video.

One of her most important contributions to the field was her observation that image processing algorithms weren't working like human vision does-- while the human eye is logarithmic, the algorithms were using linear operations to perceive and process information. So she began to develop algorithms that were modeled on the Human Visual System, giving computers the ability to “see” and evaluate images the way people do. She also created measures that enable computer systems to quantitatively evaluate images on their own in real time, allowing them to automatically produce enhanced images of the best visual quality for human viewing.

Dr. Panetta’s work has always been inspired by real-world events and improving human lives. After the attacks of 9/11, she decided to use her talents to develop efficient algorithms for image and video enhancement in real time for safety and security applications. Motivated by the 2014 AirAsia plane disaster, she began working on enhancement methods that correct for the color casting in images taken underwater, where the light tends to give images a problematic tint that makes it difficult to detect objects. Using her algorithms, search and rescue missions are now better able to see color in underwater environments.

• See articles on image enhancement ➟
• See related images ➟

If an image is enhanced, how do we know which enhancement method resulted in the best visual image for further analysis? Usually, we use human subjective evaluation to answer this question, but this isn’t feasible when we’re using a computer vision application that needs to work autonomously. Dr. Panetta’s research team has introduced state-of-the-art quantitative assessment measures that evaluate the quality of an enhanced image in a manner consistent with human visual perception. By quantifying the results of different image enhancement algorithms, they allow autonomous systems to determine the best enhancement method for a specific application.

• See articles on Quality Assessment Measures ➟
• See related images ➟
• Download ➟

Edge detection is the first step in image analysis and processing. The goal is to detect the boundaries of the objects in an image, keeping important information about the shape and structural features of those objects while removing less relevant data. This reduces the complexity of the information in the image, which reduces the computation required to further process the image. Dr. Panetta has developed edge-detection algorithms that make it easier and quicker for airport screeners to recognize and interpret objects on their X-ray screens and that help radiologists detect cancerous tissue and track the growth of tumors. They also make pattern- and face-recognition systems more efficient and effective.

• See articles on Edge Detection ➟

To further use images for biomedical and defense applications, it is often necessary to remove “noise” from the images. (“Noise” in a digital image is like “snow” on a TV.) For example, the noise in a CT scan goes up as the amount of radiation exposure goes down—and the raw image would be indiscernible to the human eye. Dr. Panetta works on algorithms that remove noise from images while preserving their details, making them less grainy, more clear, and easier to interpret.

• See articles on Noise Removal ➟
• See related images ➟

Dr. Panetta works on fine-tuning algorithms to produce the sharpest, clearest images for biomedical applications. Medical images often suffer from poor resolution and contrast. When enhancing these images, it is important to preserve all of the important data while removing irrelevant information that could be distracting, without introducing artifacts or filtering out abnormalities that could lead to false positive or false negatives.

• See articles on Biomedical Imaging ➟
• See related images ➟

Dr. Panetta works on detection algorithms for facial recognition systems. These algorithms help computer systems more quickly and accurately find and rank the photographs in a database that most closely match a test image (like a photo of a missing child or composite sketch of a suspect).

• See articles on Detection and Recognition Systems ➟
• See related images ➟

Dr. Panetta’s lab was the first at Tufts University to use UAVs and combine multiple sensors on UAVs. Her team is working to produce UAVs that can fly into disaster situations to identify and relay information about potential hazards or threats before first responders begin rescue and recovery missions. While UAVs have been used by emergency response teams in the past, they tend to be designed for one specialized task. Creating a UAV that combines video with various sensors would enable the same drone to be used in different emergency situations. The current focus is on integrating radiation sensors onto a commercial drone. Disasters involving radiation can be extremely dangerous due to the additional health risks. The use of UAVs as remote radiation detection would allow response teams to quickly access the disaster area. Responders could get to survivors faster without endangering their own lives.

• See articles on Robotic Unmanned Aerial Vehicles (UAVs) ➟
• See related media ➟
• DraganEye Video Transmission Tutorial (PDF) ➟
• DraganFly Guardian API Installation Instructions (PDF) ➟
• Automated Threat Detection for Disaster Response Teams Using UAV Platforms (PDF) [Andrew's Schweig's thesis] ➟

A few years back, Dr. Panetta, a trained opera singer, suffered from chronic laryngitis. While getting speech therapy at Massachusetts General Hospital in Boston, she encountered young women who, after battling thyroid cancer, had to use horrible sounding mechanical devices in order to talk. She wanted to create a device that would sound more human—so her team took the same approach with sound as with vision: to develop a device that sounded more human, they modeled how the human voice work to create sound. Working with speech pathologists, they created the Electrolarynx—a tubular voice prosthesis that uses signal processing to produce natural-sounding voices for people who have had their larynxes surgically removed. Her work has resulted in a patent with her students to address these problems.

• See related images ➟

Before joining the faculty at Tufts University, Dr. Panetta was a Principal Engineer for Digital Equipment Corporation. There, Dr. Panetta worked on several major processors including the Alpha Risc 21064 chip and generations of processors. Karen and her DECSIM® teammates were the first to use behavioral modeling to simulate huge chip designs and the first to be able to propagate faults through behavioral models.

Dr. Panetta’s work with concurrent fault simulation resulted in the first program, a diagnostic program, being executed on a fully simulated CPU. This breakthrough allowed computer manufacturers to be able to co-develop and test architectures before any hardware was manufactured. Karen further extended this work and created Multiple Domain Concurrent Simulation algorithms, which allowed software developers to look at a potentially exhaustive number of test scenarios to investigate catastrophic conditions that make programs and hardware fail or operate unsafely. Dr. Panetta embarked on this work as her Ph.D. dissertation after learning about the Therac-25 X-ray system failures.

• See related articles ➟

Elephant Conservation: What do Electrical Engineering Professor Dr. Karen Panetta and Elephants have in common?

Dr. Panetta is a founding member of the J2 (Jumbo to Jumbo) Elephant conservation initiative. She joins a team of Tufts University Engineers, Veterinarians and Biological Anthropologists to use technology to aid in the conservation, protection, health and reproduction. Dr. Panetta’s is using her expertise is using remoting imaging to identify and track individual elephants, and diagnose the animals’ health, including dental issues and reproductive health. She is also investigating elephant communication, including elephant’s ultra-low frequency audio communication systems and sensory devices.

She incorporates unmanned aerial vehicles (drones) with the intention of helping to monitor and gather images while ensuring minimal stress on the animals.

She is using signal processing to model elephant communications, including their audio communications and the tactile sensors on the pads of their feet, which can detect and distinguish movement miles away.

The goal of her research is to use this information to help intercept poachers and to move elephants away from both poachers and farm areas, where farmers are odds with them for destroying crops.

Karen advised the President of Malawi, Her Excellency Joyce Banda on STEM initiatives and now she is embarking on a new collaboration with the Malawi government, trying to protect Malawi’s elephant population.

Karen is currently co-teaching a course on Elephants through the Tufts University Experimental College, where students can learn more about using technology to help elephants.

The Tufts University ExCollege Course: The Jumbo Imperative: On Elephants and Elephant Conservation

Karen and her students Changing the World using technology: "Doing the Right Thing"

During her travels around the world, Karen has had the opportunity to see firsthand the need for engineers to use their skills to benefit humanity, especially for global social challenges facing women and children. Karen’s commitment to doing the “Right Thing” means that she takes on projects that may be controversial and oftentimes, serve populations that have no financial means or “voice” in society. Karen’s work brings awareness to the plights of these populations and a provides humanitarians with technology to better serve those individuals they are seek to help.

Why are Girls under the age of 4 in Tamil Nadu India more affected by Autism than boys?

Karen and her team’s work on Autism for the state of Tamil Nadu India helped NGO’s collect data on millions of Autistic children and help doctors identify root nutritional causes of their affliction. The outcome raised awareness to government leaders who then had the data and proof to justify allocating resources to address the problem. In 2011, His Excellency Abdul Kalam came to the US and thanked Karen and her team for making such an impact in the lives of these young girls and their families.

An IEEE Profile piece on the Autism research

To bring the advances of modern medical science to those members of our society who in spite of their great suffering have little access to care.

Why do 3 out of 4 babies born in Liberia die at birth?

Women in Liberia do not have access to health care education or services. Women giving birth often deliver babies without skilled medical providers or mid-wives present, thus with basic education, many deaths could be prevented. Dr. Mollica’s team uses a “train the trainer” approach to disseminating this information and provides the trainers with access to medical healthcare services for themselves and their families as an incentive for participation. Dr. Panetta’s team is developing low-cost training devices with educational material developed for women and young girls who cannot read or write. Beyond using simple audio, our approach uses storytelling including components of dance, local musical genres, and other art forms to make it appealing and easy for Liberian women to learn and disseminate this information. Female circumcision is a culturally difficult topic to discuss in Liberia. How then, can we educate women about the dangers of this practice and overcome the social pressures that perpetuate the practice? Developing technology to effectively solve this problem requires an in-depth understanding of cultures and societal roles and not just creating the hardware/software delivery platform. We partner with medical professionals, local leaders, NGOs to ensure that the technology we create is not just easy to use, but delivers information conducive to learning while respecting their culture.

Why can Human Traffickers use Technology better than the Law-Enforcement Agencies and NGOs trying to stop trafficking?

According to the U.S. State Department, approximately 800,000 people are illegally trafficked across international borders every year. Some 80% of these victims are female and half are children. Meanwhile, thousands of unaccompanied minors are seeking refuge from war-torn countries, only to be sold into slavery or used as live organ donors. Major challenges impeding efforts to stop Human Trafficking include:

• Criminals utilize up-to-date-technology to traffic victims and simultaneously subject them to both physical and psychological abuse.
• There is no single comprehensive tool/database to bridge technology, communication and information gaps regardless of whether the victim is a refugee from war, a runaway snared by drugs, or kidnapped by a stranger, relative or acquaintance.
• Human Tracking-information comes from a vast quantity of sources and languages across social media, websites, criminal databases, public transportation cameras and YouTube images/videos.
• Lack of a campaign for combating HT that comprehensively examines all three HT components simultaneously, namely using technology to help law enforcement, health and mental health organizations to improve safety and protection.

Dr. Panetta has partnered with Dr. Mollica, Harvard University Professor, and Physician at the Massachusetts General Hospital in Boston, and her long-time research partner, Dr. Sos Agaian, IEEE Fellow and Distinguished Professor at the University of Texas, San Antonio to create a Human Trafficking Combat team utilizes our collective experience to develop technologies to arm law-enforcement, health organizations and government agencies with the tools and data mining techniques to conquer Human Trafficking. Our current projects are being used in refugee camps in Greece, Syria and Lebanon.