Visionary.ai, a developer of image processing software, today announced the launch of a real-time video denoiser that improves video image quality. It can be applied to extend the operating conditions for the majority of the approximately seven billion image sensors manufactured each year. AI has been previously used for image noise reduction, but mostly for still images. The algorithms that Visionary.ai have developed are particularly efficient and sufficiently lightweight to be deployed on cost effective silicon, and to run in real time, at the edge.
"In very low light, when there are few photons for an image sensor to capture, noise is the limiting factor,” says Yoav Taeib, CTO and Co-Founder at Visionary.ai. “For human vision applications, this noise adds speckles, blurs, and distortion to images, and for machine vision it reduces the accuracy of object recognition. It is not feasible to simply capture, say, ten times more photons because that would need an image sensor and lens that is ten times bigger, driving up costs. An AI based approach, that uses the raw image data and uses a sophisticated algorithm to separate the noise from the image signal is a more effective way to extend camera performance.”
The Visionary.ai denoiser has been benchmarked against other approaches to noise reduction. The only alternative with comparable denoising performance (Restormer) required considerable computer processing power and took 212 times longer to execute, thus eliminating it as potential solution for real time vision. Furthermore, the Visionary.ai de-noising performance, does not come at the expense of image sharpness.
The new denoiser uses the image sensors raw output data, which has not been compressed or degraded by any post processing. It has been designed to work alongside the company’s other core technology, a software image signal processor to create a solid foundation for an evolving suite of image enhancement features.
“We have been speaking with some of the world’s leading manufacturers of electronic equipment,” explained Oren Debbi, CEO and co-founder, ”and from the level of interest that we are receiving, it is clear that noise, and low light performance, are serious issues which previous approaches have not been able to adequately resolve. We expect that, over the next 7 years, AI based approaches to image noise reduction and performance improvement will become the new standard in the electronics industry.”
"We are confident in the superior performance of our denoiser and want to make it as easy as possible for our customer's engineering teams to experience, evaluate and compare the performance before committing their resources. All that is required are samples of the raw signals from their existing image sensor, particularly those from poor lighting conditions, and within a week it is possible to see the potential image improvements that can be realized. “
