Oppo’s plan on perfecting photos and videos with AI: just embrace HVS (that’s not VHS!)،
As of December 31, 2023, OPPO has filed over 3,160 global AI-related patents. That might sound like a lot, and 3,160 is a pretty high number – but if you want to stay in the AI craze race, you'd better be prepared to work hard.
Oppo, unlike Apple, rejects the idea of lagging behind Samsung and Google in the AI race and has a thing or two to show.
The brand that gave the world the Oppo Find N3 (aka the OnePlus Open, aka one of our favorites and probably the best foldable phone of 2023) is at MWC and while it's certainly showing off some new devices, Oppo also shared some AI-related tools and platforms to further improve the hardware they sell.
The Oppo AI tools and platforms we're about to showcase are designed to handle something very annoying: blurry photos and videos.
As Oppo says:
A major gap we identified in the imaging capabilities of current smartphones is the ability to capture fast-moving objects with sufficient clarity. Based on our expertise and knowledge of imaging technologies, we tried to find a way to solve this problem, which resulted in a new set of AI Motion features.
.
Enter AI Motion technology.
What is EVS? What is HVS? (Also, if you are a minor: what is VHS?)
From the start: VHS was the closest thing to magic that 80s and 90s kids everywhere had in their living rooms: a black plastic oven, about the size of a laptop ( but much thicker), which played films (they came too). shaped like a black plastic brick). Ah, those dark ages of home entertainment before Netflix!
My speech is not totally off topic, as anyone who has had the honor of handling a VHS knows. questionable image quality could be obtained. VHS tapes were prone to wear and tear, and more: tracking errors, snowy picture patterns, color bleeding, ghosting, jitter, wrinkles, and more.
Of course, mobile technology has come a long way and smartphone cameras can't be accused of offering the same VHS picture problems in 2024, but they are far from perfect.
To bring them closer to perfection, Oppo, in collaboration with AlpsenTek and Qualcomm, unveiled AI Motion technology at MWC to advance the imaging capabilities of smartphones, meeting the challenge of capturing clear images of fast-moving objects .
AI Motion uses Event-Driven Vision Sensing (EVS) and Hybrid Vision Sensing (HVS) technologies, enabling high-resolution, motion blur-free photography and videography.
- Conventional image sensors
Conventional image sensors combine all motion information during a specific period of time (i.e. the exposure period) into a single image. This leads to motion blur in images and loss of valuable motion data.
- EVS (Event-Based Vision Detection)
EVS (Event-based Vision Sensing) is an imaging technology that continuously records motion information using a shutter-less mechanism. By responding only to changes in light intensity, EVS sensors can capture motion information with high temporal resolution at lower computational cost, allowing them to achieve a better balance between high frame rates and high dynamic range with low data redundancy and low power consumption. Despite its many advantages, EVS technology often lacks critical static visual information needed for many machine vision applications.
- Hybrid Vision Sensing (HVS) technology
HVS combines EVS and conventional imaging technology to simultaneously produce and process high-quality RGB images and EVS data, so that we can efficiently capture images with integrated motion information that is highly compatible with data processing. To better leverage EVS data information for improved image quality, Oppo has integrated AlpsenTek's hybrid vision sensor, ALPIX-Eiger, into smartphones and played a leading role in the development of a set of AI Motion algorithms capable of removing image blur and supporting video interpolation for high-quality images. image videography, enabling high-quality, real-time capture of moving objects at high speed. With support for the Snapdragon mobile platform and motion data captured by EVS, the AI Motion algorithm can restore more details lost due to motion in images with increased precision. Without the need to sacrifice exposure time, the solution delivers overall improvements in image quality through greater clarity and realism, whether the subject of the image or the phone itself which moves. The two photo examples below demonstrate the effect of deblurring on image quality. The AI Deblur feature helps capture even fast-moving subjects without ghosting or blurring, resulting in clearer images with sharper details.
Additionally, high frame rate video interpolation is supported, taking slow-motion videography to a new level. Unlike conventional constant time interval interpolation, Slow Motion is achieved by capturing motion information that can be used to recover motion trajectories more accurately, resulting in 960 slow motion video. fps or higher at theoretical 2K/4K resolution.