High performance FPGA architecture of guided filtering

Ren Guanghui; Wang Gangyi; Jin Yansheng

The guided filter is an edge preserving filter proposed in recent years, which can be broadly used in many image processing applications such as smoothing, detail enhancing, dehazing, etc. A high performance VLSI (Very Large Scale Integration) architecture of guided filter for single image was proposed. The architecture of the guided filter was fully pipelined, which can process images at the speed of nearly one pixel/cycle, without any off-chip memory. Moreover, a method which enable the size of the filtering window to be flexibly changed at run time was achieved. The architecture was implemented on FPGA of Altera's Cyclone Ⅲ and synthesis result shows that the requirement of logic elements and memory is acceptable for low-end FPGAs. Additional experimental results show that the proposed architecture is capable of processing video with dimension of 1 024 by 1 024 at the frame rate of 100 FPS. The architecture can be used as accelerator card for PC or front end of embedded systems for different kinds of real-time image processing tasks.

1. School of Electronics and Information Technology,Harbin Institute of Technology,Harbin 150001,China;

2. Tianjin Jinhang Institute of Physical Technology,Tianjin 300192,China

Received Date: 2012-06-11

Rev Recd Date: 2012-07-13

Publish Date: 2013-02-25

Key words:

image processing /
guided filtering /
VLSI /
FPGA

Abstract: The guided filter is an edge preserving filter proposed in recent years, which can be broadly used in many image processing applications such as smoothing, detail enhancing, dehazing, etc. A high performance VLSI (Very Large Scale Integration) architecture of guided filter for single image was proposed. The architecture of the guided filter was fully pipelined, which can process images at the speed of nearly one pixel/cycle, without any off-chip memory. Moreover, a method which enable the size of the filtering window to be flexibly changed at run time was achieved. The architecture was implemented on FPGA of Altera's Cyclone Ⅲ and synthesis result shows that the requirement of logic elements and memory is acceptable for low-end FPGAs. Additional experimental results show that the proposed architecture is capable of processing video with dimension of 1 024 by 1 024 at the frame rate of 100 FPS. The architecture can be used as accelerator card for PC or front end of embedded systems for different kinds of real-time image processing tasks.

HTML

Reference (15)

[1]
[2]	Lindenbaum M, Fischer M, Bruckstein A M. On Gabor's contribution to image enhancement[J]. Pattern Recognition, 1994, 27(1): 1-8.
[3]
[4]	Perona P, Malik J. Scale-space and edge detection using anisotropic diffusion[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1990, 12(7): 629-639.
[5]	Lagendijk R L, Biemond J, Boekee D E. Regularized iterative image restoration with ringing reduction[J]. IEEE Transactions on Acoustics, Speech and Signal Processing, 1988, 36(12): 1874-1888.
[6]
[7]
[8]	Luo Haibo, Shi Zelin, Hui Ying, et al. Real-time large window-sized 2D median filter based on multi-phased grouping and sorting network[J]. Infrared and Laser Engineering, 2008, 37(5): 935-939.(in Chinese)
[9]	Paris S, Durand F. A fast approximation of the bilateral Filter using a signal processing approach[C]//ECCV, 2006: 568-580.
[10]	罗海波, 实则林, 惠颖, 等. 基于多级分组排序网络的实时大窗口二维中值滤波器设计[J]. 红外与激光工程, 2008, 37(5): 935-939.
[11]	Porikli F. Constant time O(1) bilateral filtering[C]//IEEE Conference on Computer Vision and Pattern Recognition, 2008.
[12]
[13]	He K, Sun J, Tang X. Guided Image Filtering[C]//ECCV, 2010: 1-14.
[14]
[15]

High performance FPGA architecture of guided filtering

Abstract

References

Proportional views

通讯作者: 陈斌, bchen63@163.com

Article Metrics

Related

Proportional views