基于HLS的色彩插值算法硬件设计与实现

Hardware acceleration design and implementation of color interpolation algorithm based on HLS

为了快速、高效地在FPGA平台上实现色彩插值算法并得到高质量的彩色图像,使用Xilinx高层次综合工具HLS设计并实现一种改进的色彩插值算法。该算法融合双线性插值算法和一阶微分边缘导向插值算法完成图像色彩插值,对以R分量与B分量为中心像素点的像素使用一阶微分边缘导向插值算法插值计算缺失的颜色分量,对以G分量为中心像素点的像素使用双线性插值算法插值计算缺失的颜色分量。该算法充分考虑图像边缘插值效果与FPGA硬件资源占用量,可对Bayer格式图像进行还原,恢复全彩色图像。本文使用HLS开发,与传统FPGA开发相比缩短了开发周期,提高了开发效率。在Kodak数据集上的测试结果表明,改进的色彩插值算法的彩色峰值信噪比(CPSNR)相较于常规算法高4~6 dB,且减少了图像锯齿现象和边缘模糊等问题,本文算法占用FPGA资源较少,可实现实时的图像色彩插值。

To quickly and efficiently implement a color interpolation algorithm on the FPGA platform and get a high-quality color image,an improved color interpolation algorithm is designed and implemented using Xilinx high-level synthesis tool HLS.The algorithm combines the bilinear interpolation algorithm and first-order differential edge oriented interpolation algorithm to complete image color interpolation.The first-order differential edge oriented interpolation algorithm is used to calculate the missing color components for the pixels with R component and B component as the center pixel points,and the bilinear interpolation algorithm is used to calculate the missing color components for the pixels with G component as the center pixel points.The algorithm fully considers image edge interpolation and FPGA hardware resources,which can restore a Bayer format image and restore full-color image.Compared with the traditional FPGA development,this paper uses HLS to shorten the development cycle and improve development efficiency.The experimental results on the Kodak dataset show that the color peak signal-to-noise ratio(CPSNR)of the improved color interpolation algorithm is 4~6 dB higher than that of the conventional algorithm,and the image aliasing and edge blur are reduced.This algorithm takes up less FPGA resources and can realize real-time image color interpolation.