论文部分内容阅读
自动聚焦是数码设备、计算机视觉中的一项关键技术。自动聚焦过程中,聚焦的准确性和抗噪声性能至关重要。以高频分量作为度量的聚集评价函数具有灵敏性高、聚焦准确的优点,适用于实时系统,但是对噪声十分敏感,受噪声污染时可能导致聚焦失败。因此,提出了一种具有噪声稳健性的高频分量自动聚焦评价函数。该函数通过小波多分辨力分析提取高频分量,利用了信号的每个子带的小波系数存在一定相关性,而噪声不存在这样的相关性的特点,设定高频子带阈值,认为低于阈值的系数是噪声的贡献,大致分离图像信号与噪声信号,从而将其滤除。经过大量的实验,证明提出的方法具有单峰性好、灵敏度高等优点,特别是在抗噪声性能方面有很大提高。
Autofocus is a key technology in digital equipment and computer vision. Focusing accuracy and anti-noise performance are crucial during auto-focus. The high-frequency component of the aggregate evaluation function has the advantages of high sensitivity and accurate focusing, suitable for real-time systems, but is very sensitive to noise and may cause focusing failure when contaminated by noise. Therefore, a high-frequency component auto-focusing evaluation function with noise robustness is proposed. The function extracts the high frequency components by wavelet multiresolution analysis, utilizes the correlation of the wavelet coefficient of each subband of the signal, but the noise does not exist such a correlation. The high frequency subband threshold is set to be lower than The coefficient of the threshold is the contribution of the noise, which roughly separates the image signal from the noise signal so that it is filtered out. After a large number of experiments, the proposed method has the advantages of good monomodal, high sensitivity, especially in the anti-noise performance has greatly improved.