基于小波的磨粒图像融合方法研究

Research on Fusion Method of Wear Particle Image Based on Wavelet

针对油液磨粒检测需要,基于彩色空间变换和小波分析理论,研究磨粒多聚焦图像融合方法。该方法采用邻域梯度特征因子取大法来选取多聚焦图像清晰区域的低频系数,以进行低频融合系数的确定;采用建立新系数带以及系数取大法来进行高频融合小波系数的选取。分析小波基和分解层数对磨粒多聚焦图像融合效果的影响,结果表明,在分解层数为5时,采用db4、sym4、bior2.4小波基进行磨粒多聚焦图像融合均可获得较好的融合效果。

For the demand of wear particle detection, wear particle multi-focus image fusion method was studied based on color space conversion and wavelet analysis theory. The low frequency fusion coefficients from the clear regions were selected based on the characteristic factor of neighborhood gradient. The high frequency fusion coefficients were selected based on absolute value maximum by establishing new high frequency sub-band. The effects of wavelet base and decomposition layer on wear particle image fusion were studied respectively. The results show that better fusion effect can be obtained by using the wavelet base of db4, sym4 or bior2. 4 when the decomposition layer is five.