Abstract: The segmentation of motive impurity in liquid plays a key role in automatic detecting system. A fuzzy adap- five threshold segmentation algorithm of visual impurity in liquid based on spatial block variance and its gradient was proposed. At first, the two difference images were partitioned into 5 X 5 blocks, and then the product of the homologous block variance and its gradient of the two difference image were chosen as the block feature. After that, an adaptive threshold was used to segment the image. To be adapted to the change of the block grey value,a fuzzy inference method was adopted to adjust the threshold adaptively according to the block variance and it s gradient. hhe experimental and factual testing results show that the proposed algorithm can meet the demands of visual impurity in liquid in real-time detecting, and that it is a practicable and effective image segmentation method.

Key words: Impurity in liquid,Imagc scgmcntation,Diffcrcncc image variance, Variance gradicnt,Fuzzy adaptive threshold