为了实现大面阵高帧频CMOS相机输出的Bayer格式图像的实时精准还原, 恢复其真实色彩, 减少整幅图像(特别是边缘图像)的色彩失真, 提出一个既能够准确还原图像色彩又能够尽量减少复杂性的算法。首先对目标图像进行预处理, 即将图像进行扩大, 整幅图像采用统一算法进行色彩还原, 以减少边缘处像素单独还原所增加的计算量。而后利用改进的梯度插值, 通过计算一个5×5邻域内的水平梯度边缘值和竖直梯度边缘值, 统计其两个方向边缘值的和来对边界方向进行仲裁。最后利用G分量插值结果对其他两分量进行还原, 从而大大减小了传统算法所引入的色彩失真。用MSE、PSNR两种方法对本文提出的改进算法进行评估, 结果表明本方法能够改善图像色彩还原程度、锐化图像边缘、提升计算效率、减少色彩失真。该方法可以实现12k×5k@10 f/s的图像数据实时色彩还原。本算法不仅色彩还原优于传统算法, 而且适合使用FPGA实现高速图像数据的实时色彩还原。

To achieve the accurate real-time restoration of Bayer image from the output of large area array and high frame CMOS camera, restore its true color and reduce the color distortion of the whole image (especially the edge image), this paper proposes an algorithm that can accurately restore image colors as well as minimize complexity of algorithm. Firstly, the algorithm preprocesses the target image, that is, expands the image and use a unified algorithm to restore the entire image to reduce the amount of computation of channel of pixel at the image edge. Then the improved gradient interpolation is used to calculate the horizontal edge gradient value and vertical edge gradient value in a 5×5 neighborhood. Then, the direction of border is arbitrated by counting the sum of the two direction gradient values. Finally, the interpolation of G component is used to restore the other two components, which can greatly reduce the distortion of color introduced by the conventional algorithm. The proposed algorithm is evaluated by using MSE and PSNR methods. The proposed method can improve the color reproduction of images, sharpen the edges of images, improve the computational efficiency, and reduce the color distortion. This method can achieve 12k×5k @ 10 f/s image data real-time color restoration. This algorithm is not only superior to the traditional algorithm, but also beneficial to the use of FPGA to realize the real-time color restoration of high-speed image data.