基于聚焦曲面的快速聚焦算法的研究

Research on Fast Focusing Algorithms Based on Focusing Surface

在大倍率监控摄像机中,聚焦速度对摄像机有着重要的影响。为了提高聚焦速度和聚焦精度,通过对聚焦曲线的分析从而获得聚焦曲面,利用聚焦曲面,在变倍或者移动的同时进行跟焦处理,结合数字图像处理的方法,可以快速实现聚焦。为减小计算量,对分辨率为3840×2160的输出数据进行抽稀放缩,结合放缩后图像特点,使用快速DCT变换进行清晰度评价,从而实现局部小范围精确聚集。基于4K分辨率的33倍光学变焦网络摄像机对提出的算法进行验证,实验结果表明在保证画面清晰的跟焦处理结束后3.7秒内即可完成聚焦。

Focusing speed plays an important role in high power surveillance cameras. In order to improve the focusing speed and focusing accuracy, the focusing surface is obtained by analyzing the focusing curve. Focusing can be done while zooming or moving by using this focusing surface. With the method of digital image processing, focusing can be realized quickly. To reduce the computational complexity, the output image with resolution of 3840×2160 is scaled and diluted. With the characteristics of scaled image, the fast DCT transformation is used to evaluate the scaled image, so that local small-scale accurate focusing can be achieved. The proposed algorithm is validated by a 33x optical zoom network camera with 4K resolution. The experimental results show that the focusing can be completed within 3.7 seconds after fo? cus tracking under the condition that the picture is clear.