为了提高自动调焦算法的性能，对调焦评价函数和调焦搜索算法进行了研究。在分析人类视觉系统(HVS)特性研究成果的基础上，提出了一种基于HVS加权的小波调焦评价函数。根据视觉多通道特性和视觉敏感度带通特性，对不同方向、不同空间频带的小波高频系数分别赋予不同的权值。为了克服爬山法搜索速度慢的缺点，提出了一种基于SOM神经网络的搜索算法，采用训练完成的SOM神经网络预测镜头的最佳聚焦位置。实验结果表明，提出的小波调焦评价函数得到了比传统小波调焦评价函数更优的调焦特性曲线;采用本文的搜索算法平均仅需要采集处理73幅图像就能找到最佳聚焦位置，与基于全搜索的爬山法相比，节省了大量的搜索时间。说明提出的调焦方法不但可以得到符合人眼视觉的调焦效果，还可以实现快速自动调焦。

In order to improve the performance of the autofocusing algorithm, focus measure function and focus search algorithm are studied. On the basis of analyzing the research results of the characte-ristics of the human visual system (HVS), a weighted wavelet focus measure function is proposed based on HVS. According to the visual multichannel characteristics and the visual sensitivity bandpass characteristics, the different weights of the high frequency coefficients are set in different orientations and different spatial frequencies. To overcome the disadvantage of slow search speed of the mountain climbing searching algorithm, an improved searching algorithm is presented based on selforganizing map (SOM) neutral network. The welltrained SOM neutral network is applied to predicting the best focused lens position. The experimental results demonstrate that the characteristic of the proposed wavelet focus measure curve is better than that of traditional wavelet focus measure curve. Only 73 images need to be collected and processed averagely in the process of finding the best focused lens position by improved searching algorithm. Compared with the full search method, the proposed approach can reduce the search time greatly. It is concluded that this autofocusing method can not only obtain the focusing effects in accordance with the human visual system, but also achieve autofocusing rapidly.