Low latency image processing IP for ADAS / AD applications
Modern automobiles contain a growing number of cameras for surround view and camera based rear view systems. Omnitek provide a range of IP Cores specifically implemented for automotive image processing of surround view cameras. These include the ISP IP (to process the RAW sensor image), the HDR Tone-mapping IP (to improve image dynamic range and contrast) and the Image Stitch IP (to stitch images from different camera sources into a single composite image.
Our Image Stitch IP has been designed with the highest quality surround view systems in mind. Using a very small footprint IP, the Omnitek system is able to automatically stitch together up to 8 HD source images to create a 4K composite image for exceptional quality surround view displays.
Omnitek’s 3D Depth Map IP can automatically generate a 3D depth map from stereo cameras to estimate the distance of objects in view.
In this example, multiple RAW camera images are being processed by multiple instances of the ISP IP, enhanced by corresponding HDR Tone-mapping IP and then stitched together as a composite display by the Image Stitch IP. The Warp Processor within the Image Stitch IP will intelligently stitch together the source images and make the appropriate geometric and colour corrections to produce a composite image.
The Image Stitch IP can be configured to stitch together up to 8 camera images to produce a panoramic or 360 degree surround view display.
AI is being used in an increasing number of Driver-Assistance and Autonomous Driving applications, including the detection and recognition of cars, pedestrians, road signs etc. Omnitek’s industry-leading Deep Learning Processing Unit (DPU) can easily be optimised for such applications and seamlessly integrated with the above IP to create complete systems for ADAS or AD.