基于梯度方向的反马赛克算法

目前数字图像采集系统中采用的反马赛克算法存在插值伪像严重、算法计算量大等问题.为了解决上述问题，本文提出了一种基于梯度方向的反马赛克算法，可提升实际应用中反马赛克处理性能.计算当前像素位置水平和竖直梯度，并据此完成G分量的插值恢复，采用双线性插值完成R分量和B分量恢复并用G分量进行校正处理.从复合峰值信噪比（CPSNR）以及计算量两方面将本算法与其他典型算法进行对比，本文算法以更低的运算量，获得优秀的反马赛克性能.这种高性价比的特点有利于其未来完成硬件电路实现，并应用于实时成像等高技术领域. 

A Demosaicking Algorithm Based on the Gradient Direction

At present, many problems exist in the anti-mosaic algorithm used in digital image acquisition system, such as serious interpolation pseudo-image and large computation. In order to solve the above problems, an anti-mosaic algorithm based on gradient direction was proposed to improve the demosaicking performance in practical application. Firstly, the horizontal and vertical gradients of the current pixel position were computed, and the G channel was interpolated accordingly. Then, the R channel and B channel were interpolated based on bilinear algorithm with the calibration of G channel. And this algorithm was compared with other typical algorithms from the perspective of composite peak signal noise ratio (CPSNR) and computation. The comparison results show that the proposed algorithm can consume lower computation amount and achieve excellent performance of demosaicking processing. In future, with its higher cost-effective feature, it will be used in real-time imaging and other high-tech fields.