摘要
基于彩色滤波阵列(CFA)的图像传感器在每个像素位置获得三原色红(R)、绿(G)和蓝(B)中的一种分量,其缺失分量需要根据周围像素插值得到。目前提出的许多种插值算法,绝大部分采用Bayer排列模式。本文在色差恒定假设基础上,提出一种基于双边滤波器的自适应Bayer模型插值算法,对G通道的估计采用自适应滤波器进行插值,对R和B通道的插值采用双边滤波器。算法利用待插值像素与不同距离像素相关性不同的思想,根据图像边缘自适应设定滤波模板,能较准确估计G、B和R、G通道之间的色差值。实验结果表明,对比多尺度色差梯度算法和边缘强度滤波等算法,插值后的图像不仅主观视觉,且客观评价指标(彩色峰值信噪比,CPSNR)均优于这些算法。
Single image sensor based on color filter array （CFA） only samples one color at each pixel co- ordinate. Reconstructing the full-color image,missing colors need to be interpolated according to neighbor colors. Many methods have been developed to interpolate missing pixels in the CFA pattern, and Bayer CFA pattern is the most commonly used structure. In this paper, we propose an adaptive algorithm using bilateral filter based on the criterion of color difference constant. This algorithm uses adaptive filter for green channel interpolation and bilateral filter for red and blue channel interpolation. The proposed algo- rithm exploits the correlative information depending on both the color of neighbor pixels and the relative distance between neighbor pixels within an image. It can adaptively choose ＇the filter template and im- prove the color difference between blue and green channels or red and green channels. Experimental re- sults show that it outperforms the multiscale gradients and other recent interpolation methods in terms of both subjective vision and objective assessment （color peak signal noise ratio）.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2014年第11期2067-2073,共7页
Journal of Optoelectronics·Laser
基金
国家自然科学基金(61373151)
上海市自然科学基金(13ZR1415000)资助项目
