Imaging technology company Immervision said it is now able to offer real-time video distortion correction software algorithms to remove stretched bodies, change lines, objects and face proportions and adjust scenes in real-time on images and videos captured on smartphones.
The company will license the new algorithms directly to mobile OEMs through its distribution partner and lead investor CEVA. In an interview with EE Times, vice president and co-founder of Immervision Patrice Roulet Fontani said this new real-time, on-the-fly distortion correction wasn’t really available before. “Some things were there, but to be able to run on a portable device without draining the battery, and providing the appropriate correction, you didn’t have that before.”
He explained the background. “Smartphones are having wider screens, but up to now, whenever you look at a video, it’s still in a 4 by 3 type of ratio. If you want to fill the screen you need to process the image to crop and then split and then increase the size of the image on your sensor. Now with cinematographic and wide-angle lenses, they allow you now to fill the screen, but not without having some artefacts or some distortions. Correcting the picture is one thing, because you can snap the picture, maybe get half a second to process it and modify it. But with a 30-120 fps video, that’s another story.”
“To bring that distortion correction, we have introduced a new algorithm that not only corrects the distortion in terms of one setting, but you can also have multiple types of distortion correction, depending on the situation. And it can be done in real time, correcting areas for example where there are people, we might want to have the people look more normal. Where you have panoramas, you maybe want to provide a feeling of being a wide, immense area.”
“What you might see in existing devices today is that they do not correct the distortion at all. There is distortion correction for images, and videos, but they are focused very much on the straight line. And they correct only one distortion correction feed that feeds all the applications scenarios. What we notice with this is, you cannot have one processing that corrects the straight line, and at the same time preserve the body or face proportions. So, in most smartphones today, the body and face proportions as you go towards the edge of the FOV [field of view] are very much distorted. So, they are stretched where the face or the body appear wider than they are.”
“Instead of having one processing feed for all scenarios, we provide smartphone OEMs different video distortion correction processing, depending on the scenarios. So, if you want to capture a landscape, it would be a different processing to if you have a group of people or portrait image. We developed video distortion correction plus what we call face protection, to make the faces appear as they are in reality and not distorted and not stretched. In addition, this process is performed in real time. Our commitment to our customer is to enable what you see is what you get.”
In many newer phones, the wider field of view creates more apparent distortion. Immervision said its new software algorithms help remove stretched bodies, change line, object and face proportions in real-time adjusting to the scene, and reduce editing time, all in one solution. The software provides different levels of correction, varied projections and real-time adaptation to the scene. This gives smartphone OEMs the capability to offer different options to its users – for example, to leave preset configurations as they are, or be able to fully customize, or let end-users make the decision, as well as allow phone orientation to dictate or leverage machine learning to control the experience.
Fontani said being able to enable distortion correction in real-time is important. “Because when you are filming, there are two things happening. One you are recording, or you are broadcasting, that needs to be done as it goes. You cannot record and then process after.”
Immervision said with the skyrocketing demand for video recording on phones, software is needed to create the ideal framing for high-quality videos. Having worked with its customers for 20 years, it said its hardware and software experts have been working closely with OEMs to fine tune the algorithms to meet specific needs. The new algorithms, it added, help deliver “20/20 vision” in a range of smartphones, correcting video distortions in real-time to improve the captured to display experience, without compromising the field of view.
Adding an off-the-shelf 125 degrees wide-angle lens
In a second announcement, Immervision also introduced what it claims to be the highest-performing off-the-shelf 125-degree wide-angle lens, pre-configured on popular sensors without need for camera customization, and aimed at tier two and tier three mobile OEMs.
Wide-angle lens reduces the number of cameras and costs, while improving image quality for landscape, group, and close-up pictures and videos taken with mobile phones. While the majority of tier one OEMs have wide-angle lens in their phones, tier two and tier three mobile brands have yet to adopt. Immervision’s technology has been preconfigured on popular sensors including Sony, Omnivision and Samsung.
“Tier two and tier three mobile OEMs are challenged with delivering the same level of image quality of leading-edge mobile phones,” said Immervision’s executive vice president, operations and chief commercial officer, Alessandro Gasparini. “Immervision’s 125 wide-angle lens integrates all the industry-leading metrics into one lens. In the past, these best-in-class metrics were found in multiple phones. Now they are available in one lens with ready-to-use software, reducing camera customization and integration time, and providing the same image quality as in well-recognized mobile phone brands.”
“This is a new optical design in which we achieved the best balance between different parameters to reach the widest possible FOV that doesn’t compromise the f-number, the relative illumination, and without any drop in resolution up to the edge, up to 21 megapixels. It will allow tier two and tier three phone vendors to be able to offer a complete experience going from wide angle to traditional imaging with one single camera. In other words, having two different features on the same camera.”
He added, “What we are doing is making this technology accessible, since it’s usually only in flagship big brand phones. This will be the best lens in terms of performance achieved at wide angle.”
The post Distortion Correction Algorithms Adapt Video in Real-Time appeared first on EETimes.