ARM on Monday unveiled the Mali-C71, its first image signal processor (ISP) built from the ground up for the automotive market.
ISPs process raw pixel data so it can be displayed to a driver or used for further processing by computer vision algorithms. ARM’s ISPs already offer high dynamic range, noise reduction and other features for devices like smartphones and videoconferencing systems, but cars require higher performance ISPs.
“The reason people add cameras to cars is to make the car safer,” explained Richard York, ARM’s senior director of ARM’s embedded segment, “either to make it capable of autonomously driving, to make the car aware of its surroundings, or to simply better enable visibility for the driver.”
The Mali-C71 is the first development to come out of ARM’s 2016 acquisition of Apical, which gave ARM the capabilities in imaging and computer vision necessary to meet the automotive market’s needs. It’s a market with major potential: In 2016, there were just over 50 million camera modules shipped in the automotive market, according to Strategy Analytics. By 2023, that number should surpass 200 million, with cameras used for rear-facing parking, automatic breaking, pedestrian detection, to replace rearview mirrors and so forth. By 2023, the average mid-range car is expected to have three cameras, while luxury sedans will be approaching 10 cameras each.
The Mali-C71 fits in with ARM’s agenda of moving beyond the mobile market, particularly into automotive.
To provide clear display images for drivers — say, for instance, if they’re looking at an image that’s prelaced their rearview mirror — the Mali-C71 offers ultra-wide dynamic range (UWDR) up to 24 stops. The dynamic range is the difference in brightness between the darkest part of an image and the lightest part. At 24 stops, the ISP far outpaces even the best DSLR cameras (which have about 15 stops of dynamic range) and in some cases captures more detail than the human eye.
The images it produces, with features like low latency and advanced error detection, also meet the needs of computer vision applications. For instance, in the image below, a computer vision application may not recognize the picture on the left as a car. The image on the right shows how the Mali-C71 manages highlights in an image to make it more clear.
Meanwhile, the Mali-C71 enables ARM partners to process images from all of a car’s cameras within just one SoC. Previously, one ISP would be needed for each camera. Yet as the number of cameras in cars increases, manufacturers are trending towards building a centralized system. Ultimately, combining all of the data collected from different cameras should create a safer system.
“In simple terms, we make it as easy as possible for our silicon partners to build the ISP into a system that cares about safety to the highest levels possible,” York said.