摘要
针对CCD分辨率高、灰度位数高会导致测量图像数据量大,严重影响处理速度,而一般降维预处理又容易丢失目标细节的问题,基于LVQ神经网络和模糊C-均值聚类技术,提出了一种新的智能降维方法。首先,对测量图像中的目标和背景分别按不同类数进行聚类;然后,完成神经网络的训练,实现对图像进行均匀降维。实验结果表明,该方法在不影响目标信息的基础上,可大大地降低测量图像的数据量。
With the improvement of the resolution and grey bit of CCDs, the mass of the measured image data will be increased and the data processing speed should be affected. If a common dimension reduction method is used, some details of the target in an image will be lost easily. In this paper, a new intelligent dimension reduction method based on LVQ nueral network and fuzzy c-means clustering technology is proposed. First, the target and background in the measured image are clustered respectively in different clusters; then, the nueral network is trained to reduce the dimension of the image evenly. The experimental result shows that this method can greatly reduce the quantity of the image data to be processed without any influence on the target information.
出处
《红外》
CAS
2007年第10期14-17,共4页
Infrared
