Combining the latest release of the CEVA-CV computer vision library with CEVA’s Android Multimedia Framework (AMF) eases the integration of CEVA-CV functions with any Android-based application processor for mobile, automotive, surveillance and consumer applications and the Internet of Things. This, in turn, significantly increases performance and substantially lowers the power consumption of CV-enabled devices by offloading all computer vision processing from the CPU or GPU.
New functions added in the latest CEVA-CV release include feature detection kernels and object recognition algorithms such as Harris Corner, Hough Transform, Integral Sum, Fast, LBP, SURF, HOG, SVM, and ORB detection and matching. These are commonly used in augmented reality applications for smartphones, tablets, wearable devices, Natural User Interface (NUI), surveillance and Advanced Driver Assistance Systems (ADAS) applications. New optical flow kernels include KLT and Block Matching, which are used for motion detection and object tracking required in camera-enabled devices to implement applications such as digital video stabilization, augmented reality and gesture recognition. CEVA-CV now also includes kernels required by The Khronos Group’s OpenVX 1.0 specification, which is set to become the key standard for cross-platform acceleration of computer vision applications and libraries. This brings the number of functions in the library up to 750.
“The CEVA-MM3101 is the industry’s most proven and widely adopted platform for computer vision, with multiple tier one semiconductor and OEM licensees already incorporating it into their next-generation products and leveraging the CEVA-CV functions to develop their applications,” said Erez Bar-Niv, CTO of CEVA. “The addition of more than 250 new computer vision functions has been widely acclaimed by the growing CEVA-MM3101 developer’s community, enabling them to develop the most advanced CV applications while reducing the design complexity and power consumption for CV-enabled devices.”