覆盖在传感器表面的彩色滤波阵列(CFA)是CCD/CMOS传感器相机获取图像信息的关键，经过 CFA 后每个像素点只能获得红、绿、蓝三种颜色中的一种颜色分量。另外缺少的两种颜色分量，需要通过周围像素的值进行插值获得。针对现有插值方法出现的复杂度高、插值还原图像颜色失真和边缘拉链效应问题，设计了一种基于色差空间的低复杂度Bayer图像插值算法。第一步利用5×5模板内的像素值判断插值方向，并用颜色之间的相关性插值获得绿色分量。在分析红、绿、蓝三种颜色相关性的基础上，利用色差原理，先恢复出缺失的绿色分量，再利用恢复出的绿色分量的高频部分替换红（蓝）分量的办法，插值缺失的蓝色和红色分量。利用Matlab程序，对5幅标准测试图和CE5T1试验飞行器自主相机拍摄的图片进行了仿真实验，比较了现有几种插值算法的复杂度和PSNR值。结果显示，文章提出的算法在主观视觉和客观评价方面都具有良好的表现。同时，算法复杂度低，适合硬件实现和图像的实时处理。%A single CCD/CMOS sensor camera to capture image is based on the Color Filter Array covered on the sensor, when the light pass through the CFA, each pixel can get only one color component of three-primary colors(red, green, blue). In order to recover the other two missing color components ,they must be obtained by interpolation. Focusing on the complexity of the existing interpolation methods, color distortion and zipper effect of the interpolation image, this paper designs a low Complexity Interpolation Algorithm for Bayer Image Based on Color Difference Space. The first step is to estimate interpolation direction taking advantage of the pixels in 5×5 template and use the correlations between green and red or between green and blue to interpolate green components. Based on the analysis of the frequency correlations between color components and by using the color difference principle, we use the high frequency part of the green component to interpolate to get the missed red and blue components. Through Matlab we carry out simulation experiments based on the images captured by our CE5T1 camera, and compare several existing interpolation algorithms with the proposed algorithm. The results show that the proposed algorithm performs well both in visual and numerical aspects. At the same time the proposed algorithm has a low complexity, suitable for hardware implementation and image real-time processing.
展开▼
