改进的联合去马赛克与放大算法及其硬件实现

Improved joint demosaicking-zooming algorithm and its hardware design

摘要:

为了从单CCD/CMOS图像探测器相机拍摄的Bayer 格式图像恢复出更高分辨率全彩色图像, 设计了一种改进的联合去马赛克与放大算法。首先,估计原始Bayer格式图像中丢失的绿色 通道信息, 根据颜色梯度的比较结果,将水平或垂直方向的绿色通道信息作为彩色图像的 G分量,并根据彩色图 像颜色通道之间高度相关的特性,计算颜色差R-G 和B-G；然后,将去马赛克过程中的梯度方法继续用于确定丢失的绿色像素,放大绿色通道 ；最后, 对颜色差值图像R-G和B-G进行放大,即可恢复出红色和蓝色通道 信息,得到放大的全彩色图 像。试验结果表明,算法恢复出的高分辨率彩色图像的峰值信噪比(PSNR)有较明显提 升,且主观目视清晰、边缘锐 利,特别是硬件实现方面能够做到实时处理。目前,本文算法已经在Xilinx公 司的FPGA上实现,并用于工程实践。满足了项目实际需要。

Abstract:

In order to obtain higher resolution full color image from Bayer CFA structure image directly,this paper presents a new joint demosaicking and zooming algorithm for digital camera equipped with a single CCD/CMOS sensor.We firstly extract GH and GV components of each pixel from m osaicking image through horizontal and vertical interpolations,respectively,and then we calculate and compare the hori zontal chrominance gradient and the vertical one for choosing the missing G from GH and GV.By exploiting the hig h spectral-spatial correlations between color channels,we compute the two color difference images R-G and B-G to reduce the demosaicking and zooming errors.Next,we obtain the horizontal GzoomH and vertical G zoomV of each pixel,and also choose the missing zoomed G^zoom by calculating and comparing the chrominance gradient .Finally,the enlarged red and blue channels are obtained by adding the corresponding enlarged R-G and B-G color d ifference images to the enlarged green channel.Experimental results prove that the proposed algorithm outperfor ms some state-of-the-art demosaicking-zooming methods in both peak signal-to-noise ratio (PSNR) measur e and visual quality.Furthermore,computational cost of the algorithm is low and it can be hardware processed in real time.Now,the proposed a lgorithm is implemented on Xilinx FPGA and used in camera system,and the perfect restoration quality and ef fectiveness meet the requirement of system design.