To avoid noticeable seams between texture patches and minimize the stretch and distortion of the pattern when tiling a texture on surfaces

several texture synthesis methods have been proposed in the past few years

most of them however are time consuming and suitable only for part of textures. This paper presents a new approach for texture synthesis. The algorithm first selects multiple seeds from the sample image and puts them in the synthesized images as initial matching points

those points can be a line

a few lines or rectangle

the algorithm then searches for new matching points among the neighbors of the initial matching points along a spiral path and continues this process in a recursive manner until the synthesizing image is fully filled. Compared with the traditional approaches

the synthesizing process of our technique is dramatically speed up and reach to 360 quantitative grade

and the synthesizing speed can be accelerated further in the luminance space while retaining almost the same result as the image obtained in RGB space. In addition

our method can cope with the variation of the texture along different directions with ease. Experiments show that synthesizing results are quite satisfactory.