摘要
本文根据被测图像纹理背景和目标区域在小波变换域内图像能量的分布不同 ,以及经过多尺度分解后在各子空间信息分布的不同 ,运用与背景纹理的方向无关 ,对目标的变化具有旋转、平移以及尺度不变性的能量特征作为检测目标的依据。为了有效地提高目标边界的定位精度 ,抑制背景噪声 ,本文提出了能量特征图和基于多尺度分析的边缘特征图相融合的特征融合方法进行目标检测。
Based on the energy distribution difference between targets and background and on the information distribution difference in wavelet field after multiscale decomposing, the energy feature is chosen as one of the detection basis. It brings the independence of background texture direction and the invariance of rotation and translation for the changes of targets. In order to restrain the noise of background, improve the edge precision and probability of detection, the fusion method of energy feature and multiscale-based edge feature is proposed. The results of experiments prove that this method has relative adaptability.
出处
《光学精密工程》
EI
CAS
CSCD
2000年第5期421-424,共4页
Optics and Precision Engineering
