Abstract: To deal with the problem of confused display in 3D visualization of a flow field, an adaptive visualization method for flow fields is put forward. Due to the complicated feature in flow field, a fuzzy support vector machine is applied to describe and cluster the flow field features. To resolve the overlapping effects caused by excessive noise or omission of details caused when noise is removed, an improved line integral convolution algorithm is also presented. In the improved algorithm a fuzzy membership function obtained by fuzzy support vector machine generates sparse noise adaptively Experimental results illustrate the efficiency of the method.