摘要
传统巴特沃斯滤波器无法对不同图像自适应调整截止频率。针对此问题,提出频率自适应二维巴特沃斯滤波器,通过高频能量在图像总能量中所占比值,自适应地调整截止频率。所采用色彩空间为色相饱和度(Hue Saturation Intensity,HSI),数据格式由浮点数取代定点数,实现高速、高精度的2D-快速傅里叶变换(Fast Fourier Transform,FFT)和2D-快速傅里叶逆变换(Inverse Fast Fourier Transform,IFFT)。深度考虑Xilinx 7系列现场可编辑门阵列(Field Programmable Gate Array,FPGA)硬件底层结构,在xc7a100tfgg484-3芯片上实现该滤波器,并在此基础上实现图像细节增强,增强后图像色彩无明显失真,图像细节更为丰富。经硬件仿真测试,分辨率为512×512的图像完成一次2D-FFT所需时间为6548.892μs,与软件计算频谱结构相似度高达99.9998%。整个设计在资源、功耗和性能之间进行了权衡。
Traditional Butterworth filter cannot adaptively adjust the cutoff frequency for different images,in view of this problem,the proposed frequency adaptive two-dimensional Butterworth filter,through the high frequency band energy in the total image energy ratio,adaptively adjust the cutoff frequency.The color space used is Hue-Saturation-Intensity(HSI),and the data format replaces the fixed-point number with a floating-point number,achieving high-speed,high-precision 2DFast Fourier Transform(FFT)and 2D-Inverse Fast Fourier Transform(IFFT).Deeply considering the underlying structure of Xilinx 7 series Field Programmable Gate Array(FPGA)hardware,the filter is implemented on the xc7a100tfgg484-3 chip,and on this basis,the image details are enhanced,and the image color is colorless and the image details are richer.The hardware simulation results show that the image with a resolution of 512×512 takes 6548.892μs to complete a 2D-FFT,which is 99.9998%similar to the spectrum structure calculated by software.The entire design trade-offs between resources,power consumption,and performance.
作者
韩玉鑫
王晓凯
HAN Yuxin;WANG Xiaokai(School of Physics and Electronic Engineering,Shanxi University,Taiyuan 030006,China)
出处
《微电子学与计算机》
2024年第1期83-92,共10页
Microelectronics & Computer
基金
山西省重点研发计划(高新技术领域)(201803D121102)。