Most of existing importance sampling methods for direct illumination exploit importance of illumination and surface BRDF. Without taking the visibility into consideration, they can not adaptively adjust the number of samples for each pixel during the sampling process. As a result, these methods tend to produce images with noise in partially occluded regions. In this paper, we introduce an incremental wavelet importance sampling approach, in which the visibility information is used to determine the number of samples at run time. For this purpose, we present a perceptual-based variance that is computed from visibility of samples. In the sampling process, the Halton sample points are incrementally warped for each pixel until the variance of warped samples converges. We demonstrate that our method is more efficient than existing importance sampling approaches.