Named as one of CB Insights
100 Most Promising AI Startups
5 Image Processing Trends to Watch in 2023
2023 will see affordable high-quality video spread across many sectors, reflecting both higher customer expectations and the emerging economies of scale of core technologies.
Due to the higher availability of more cost-efficient high resolution image sensors and AI accelerators, new imaging technologies will enhance the video capabilities across a wide range of markets including healthcare, medicine, mobile and the automotive industry, in short, wherever cameras are used.
Here are five key technology trends that will boost image quality in 2023.
1. More choice in mid performance AI chips – Today’s neural network chips are generally focused on high performance points or are components of highly complex specialized SOCs as used by big five smartphone manufacturers. There is a growing awareness of the need for lower cost, more power efficient AI chips for less demanding workloads such as video image enhancement, wake-word processing, or simple inferences. We are seeing the beginning of a positive spiral where available silicon enables more applications, which in turn drives the production of more chips that enable more applications.
2. Surveillance systems become “Software Defined” -In 2023 new models of smart security cameras will move to high-definition, high-resolution, megapixel imagery, combined with on-board AI image processing and perception capabilities. Image quality will be dramatically improved, and the data bandwidth costs reduced. For existing video camera installations, which are costly to change, retrofit solutions employing AI at the edge will become available to improve image quality.
3. Next generation video conferencing – The world has already become dependent on Work from Home and Work from Anywhere routines. There is a realization that this is not a temporary stopgap during the pandemic but a permanent part of our professional and personal lives. Companies and individuals are recognizing the value in upgrading audio and video quality to reduce “zoom fatigue” and to appear more professional. A sharp and clear image is becoming just as important as appearing on the screen with a haircut and an ironed shirt. New image processing capabilities will be used for improved lighting, opaque backgrounds, viewing angles, eye contact etc.
4. IoT marries image processing - IoT, where sensors connected to the internet interact and collaborate to improve business efficiency, will grow exponentially enabling enterprises to cut waste, reduce manufacturing costs, as well as increase the quality of goods and services. These systems will utilize enhanced image processing sensors to monitor the status of people, objects, and processes with higher speed and accuracy. By combining sharper brighter images enhanced in real time with the framework of the IoT network there will be a flurry of new applications to improve traffic management, precision farming, security, manufacturing, driver assistance, surgery, and other processes.
5. Driver Monitoring Systems become mainstream - 2023 will be the year where driver monitoring systems will become a mass market feature of new vehicles. Cameras will sense when drivers are groggy, trigger an alert, and if there is no response, brakes will be applied. These systems will use advanced AI-powered image processing to operate at night, which is when drivers are the most likely to be tired. Legislation has just come into effect in the EU for new car models, and regulators in China and the USA are actively investigating the potential of driver monitoring systems to reduce traffic collisions.
This year, advanced software-based imaging processing will provide more accurate and up to date video to improve health, safety, and business efficiency. IoT applications that rely on image processing sensors will become more effective, AI accelerators that optimize image quality will become more available, and the move towards software-based ISPs will accelerate innovation.
As AI algorithms improve their ability to learn over time, the quality of image processing will continue to improve. Device manufacturers will recognize that improving software is the best strategy for improving image quality, and software-based ISPs powered by AI accelerators will continue to raise the bar of the power of vision-based applications.