This paper presents an efficient algorithm for texture synthesis using tree-structured vector quantization (TSVQ). The algorithm generates high-quality, tileable textures by modifying random noise to resemble a given example texture. It is fast, requiring only a sample texture as input and producing results that are as good or better than previous methods, but two orders of magnitude faster. The algorithm is based on Markov Random Field (MRF) texture models and uses a deterministic searching process, which is accelerated using TSVQ. The algorithm works by first building a multiresolution pyramid from the input texture and noise. It then synthesizes the texture by comparing neighborhoods of pixels in the input texture with those in the noise. The neighborhoods are compared using an L2 norm to find the best match. The algorithm ensures that each pixel is assigned to maintain local similarity between the example texture and the synthesized image. The algorithm is efficient because it avoids explicit probability construction and sampling. It uses a multiresolution approach to handle large-scale structures and reduce computational cost. The neighborhoods are handled in a causal manner, ensuring that each neighborhood contains only already assigned pixels. This helps in generating valid results and avoiding boundary artifacts. The algorithm is also flexible and easy to use, requiring only an example texture patch. It can be applied to various applications such as image editing and temporal texture generation. The algorithm is accelerated using TSVQ, which allows for efficient nearest point searching by treating neighborhoods as points in a multidimensional space. This significantly reduces the computational cost of the synthesis process. The results show that the algorithm can generate high-quality, tileable textures quickly. It outperforms previous methods in terms of speed while maintaining or improving the quality of the results. The algorithm is also effective in constrained synthesis for image editing and temporal texture generation. The use of TSVQ allows for efficient synthesis of textures, making the algorithm suitable for real-time applications. The algorithm is easy to implement and can be extended to various applications, including texture compression and motion synthesis.This paper presents an efficient algorithm for texture synthesis using tree-structured vector quantization (TSVQ). The algorithm generates high-quality, tileable textures by modifying random noise to resemble a given example texture. It is fast, requiring only a sample texture as input and producing results that are as good or better than previous methods, but two orders of magnitude faster. The algorithm is based on Markov Random Field (MRF) texture models and uses a deterministic searching process, which is accelerated using TSVQ. The algorithm works by first building a multiresolution pyramid from the input texture and noise. It then synthesizes the texture by comparing neighborhoods of pixels in the input texture with those in the noise. The neighborhoods are compared using an L2 norm to find the best match. The algorithm ensures that each pixel is assigned to maintain local similarity between the example texture and the synthesized image. The algorithm is efficient because it avoids explicit probability construction and sampling. It uses a multiresolution approach to handle large-scale structures and reduce computational cost. The neighborhoods are handled in a causal manner, ensuring that each neighborhood contains only already assigned pixels. This helps in generating valid results and avoiding boundary artifacts. The algorithm is also flexible and easy to use, requiring only an example texture patch. It can be applied to various applications such as image editing and temporal texture generation. The algorithm is accelerated using TSVQ, which allows for efficient nearest point searching by treating neighborhoods as points in a multidimensional space. This significantly reduces the computational cost of the synthesis process. The results show that the algorithm can generate high-quality, tileable textures quickly. It outperforms previous methods in terms of speed while maintaining or improving the quality of the results. The algorithm is also effective in constrained synthesis for image editing and temporal texture generation. The use of TSVQ allows for efficient synthesis of textures, making the algorithm suitable for real-time applications. The algorithm is easy to implement and can be extended to various applications, including texture compression and motion synthesis.