VGA 91 fps Affine Motion Segmentation Processor for Video Recognition

Masayuki MIYAMA  Yoshio MATSUDA  

Publication
D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)   Vol.J97-D   No.6   pp.1122-1132
Publication Date: 2014/06/01
Online ISSN: 1881-0225
Type of Manuscript: PAPER
Category: 
Keyword: 
affine motion model,  motion model estimation,  motion segmentation,  real-time processing,  VLSI,  

Full Text(in Japanese): PDF(2.2MB)
>>Buy this Article


Summary: 
This paper proposes a VLSI architecture for VGA 91 fps motion segmentation with affine motion model. Our proposed algorithm shares image gradient calculation and motion quality calculation as preprocessing. The following calculation of gradient image moment for motion model and region labeling are processed successively pixel by pixel. Fast convergence and low computational complexity of the algorithm are confirmed by simulation. Novel pixel pipeline architecture based on the algorithm is characterized by small scale of circuit and high throughput processing. We implement the VLSI processor using 0.18 µm technology. The core area is 16.8 mm2 and the throughput is VGA 91 fps at 94 MHz operating frequency. This processor is targeted at the video recognition applications such as vehicle safety under restriction of energy consumption.