The objective of this subject can be stated as follows: given a texture sample, and synthesize a new texture that, when perceived by a human observer, appears to be generated by the same underlying process [25]. Recently, there are two general methods in this area: pixel-based synthesis method and patch-based sampling method Patch-based texture synthesis algorithms can produce reasonable results for a wide variety of texture classes. They are good at preserving global structure and can introduce artifacts along patch boundaries, but the disadvantage is that it will mismatch in the overlap region. Pixel-based synthesis algorithms, on the other hand, tend to blur out small objects while maintaining a consistent texture impression, which preserve local coherence of the source texture, but possibly fail to preserve the global structure. In this paper, a novel method of hybrid scheme for texture synthesis is proposed. By the scheme, the feature of textures in two levels is captured, At the macro-structure level, first a texton mask on the pattern from the original source texture is generated, to be employed in the synthesis as a primary guidance for the patch-based sampling. And in the micro-structure level, a blending is performed by using a pixel-based synthesis method based on the grey-level texton mask within the boundary zones of the patches. As a result, the method may not only quickly capture the macro-structure of the source texture, with little overhead from simple interactive operation, and at the same time, the micro-structure may be also nicely preserved in the synthesis by a pixel-based blending procedure. Experiments show that the solution proposed in combining patch-based sampling and pixel-based sampling method based on texton masks is highly efficient for synthesizing large amount of textures with comprehensive structures. Keywords: texture synthesis, texton masks, combination of patch-based with pixel-based