Goksel Dedeoglu

An optimized vision library approach for embedded systems

There is an ever-growing pressure to accelerate computer vision applications on embedded processors for wide-ranging equipment including mobile phones, network cameras, and automotive safety systems. Towards this goal, we propose a software library approach that eases common computational bottlenecks by optimizing over 60 low- and mid-level vision kernels. Optimized for a digital signal processor that is deployed in many embedded image & video processing systems, the library was designed for typical high-performance and low-power requirements. The algorithms are implemented in fixed-point arithmetic and support block-wise partitioning of video frames so that a direct memory access engine can efficiently move data between on-chip and external memory. We highlight the benefits of this library for a baseline video security application, which segments moving foreground objects from a static background. Benchmarks show a ten-fold acceleration over a bit-exact yet unoptimized C language implementation, creating more computational headroom to embed other vision algorithms.

Guest Editorial: Special Issue on Embedded Computer Vision

It is with great pleasure that we present this Special Issue of the Journal of Signal Processing Systems (JSPS) dedicated to Embedded Computer Vision! We are pleased to include six state-of-the-art papers from the leaders in this field, both from industry and academia, who keep pushing the embedded computer vision technology forward. While the idea for this special issue originated between the Guest Editors at one of the CVPR workshops on the same topic that we have organized, it is the work of the contributing authors that makes it a success. The papers were solicited from the workshop participants and through an open call for papers, so the initial submissions were in many ways already pre-filtered. Out of 24 submitted papers, the highly selective review process yielded the six papers included here. They cover a broad range of challenges that are encountered in practical deployment of embedded vision systems, especially when high computational performance needs meet limited resources. We present papers describing a range of novel solutions: a deep learning accelerator, a robust aerial tracking system, an FPGA-based aerial visual servoing task solution, an approach to use low-cost hardware for real-time vision, a real-time motion detector, and an image enhancement approach based on human vision. In the following we summarize their contributions: