基于Curvelet变换的浮选泡沫图像序列时空联合去噪

提出一种时空信息联合的浮选泡沫图像去噪方法。首先，将基于GSM统计建模和贝叶斯最小二乘准则的局部空间去噪方法应用到图像Curvelet域，获得基于单图像信息的Curvelet空间域最佳系数估计；然后，根据运动补偿原理和帧间子块的相关性引入帧间加权因子，通过加权处理帧间子块系数获得待处理图像时空相关的最佳去噪系数估计。结果表明，该方法能在去除噪声的同时更好地保护泡沫的细节，对于严重噪声污染的泡沫图像序列也能获得较好的处理效果。

Spatial-temporal joint for froth image sequence denoising based on Curvelet transform

An image denoising method based on temporal-spatial information fusion for flotation froth images processing
is proposed. Firstly, a kind of classic image denosing method based on local spatial statistical model of Gaussian scale
mixture(GSM) and Bayes-least square inference is applied in the image Curvelet domain to estimate the optimal coefficients
based on the spatial domain information of single image. Then, weighted impact factors of the inter-frames according to
motion compensation and similarities measurement of the sub-blocks in the adjacent frames are introduced. Consequently,
ultimate and optimal image coefficients of the froth frame to be processed are computed by weighting the adjacent frames
coefficients according to the correlation information on the spatial-temporal domain with the weighted impact factors. The
experiment and application results explicitly show that this method can achieve much better performance to protect the bubble
details of the image sequences while removing the image noise, and it also achieves good results for the seriously polluted
image sequences.