Chinese Journal of Computers   Full Text
Title	Upsampling Synthesized Texture Using Joint Bilateral Filter
Authors	XIAO Chun-Xia1),2),3) NIE Yong-Wei1) HUANG Xian-Feng4) ZHAO Yong2) PENG Qun-Sheng2)
Address	1)(School of Computer, Wuhan University, Wuhan 430072) 2)(State Key Laboratory of CAD&CG, Zhejiang University, Hangzhou 310027) 3)(State Key Laboratory of Software Engineering, Wuhan University, Wuhan 430072) 4)(State Key Laboratory of Liesmars, Wuhan University, Wuhan 430072)
Year	2009
Issue	No.2(241—251)
Abstract &   Background	Abstract This paper presents a novel technique for upsampling synthesized texture using bilateral filter. The basic idea is based on the observation: If the available high resolution exemplar texture is used as a prior to upsmaple the low resolution texture, a better high resolution result that preserving the salient structures can be obtained. A joint bilateral is used to perform the upsampling, that is, a spatial filter is applied to the low resolution synthesized texture, while a similar range filter is jointly applied on the full resolution exemplar texture. The proposed method makes computational and memory costs low to receive a high resolution image＼video texture. The experimental results show that the method can keep the salient structures as well as producing smooth upsampled texture results. Keywords texture synthesis; global optimization; bilateral filter; image upsampling Background This work was partly supported by the National Natural Science Foundation of China (No.60803081), State Key Lab of CAD&CG（No.A0808), State Key Laboratory of Software Engineering (SKLSE) (No.SKLSE2008-07-08), Ph.D. Programs Foundation of Ministry of Education of China (No.200804861038), Natural Science Foundation of Hubei Province (No.2008CDB350). Texture synthesis is a useful technique in computer graphics, it synthesize a large scale texture based on a small sample while exhibiting the same stochastic features of the exemplar texture. A variety of methods have been developed to achieve this goal. Synthesis of novel photo-realistic imagery from limited example input is of wide importance in computer graphics. Many example based synthesis approaches rely on the presence of texture. Texture refers to the class of imagery that can be categorized as a portion of an infinite pattern consisting of stochastically repeating elements. This inherent repeatability present in textures is the key behind texture synthesis techniques. These techniques generate output textures that are larger in size than the input sample but perceptually similar to it. To produce a large scale or high resolution texture is usually time-consuming, especially for video synthesis, the complexity and memory requirement sometimes make the task impossible, in this case, to upsample the low resolution synthesized texture is a feasible way. Image upsampling or enhancement is one of the most elementary image operations and is useful in many graphics applications. The commonly used upsampling techniques are trilinear interpolation or Gaussian interpolation. While these traditional methods generally assume smoothness prior for the interpolation, may suffer from blurred images. Inspired by the joint (cross) bilateral methods, We present a novel technique for upsampling synthesized texture using bilateral filter. The basic idea is based on the observation: if the available high resolution exemplar texture is used as a prior to upsmaple the low resolution texture, a better high resolution result that preserving the salient structures can be obtained. We use a joint bilateral to perform the upsampling, that is, a spatial filter is applied to the low resolution synthesized texture, while a similar range filter is jointly applied on the full resolution exemplar texture. The proposed method makes computational and memory costs low to receive a high resolution image＼video texture. Our JBU method can also be easily extended to upsample the video texture synthesized using optimization. We consider the video as a 3D video volume, in video texture synthesis, for each pixel in the synthesizing texture, we found a corresponding pixel in the low resolution texture, and its neighboring points, and then we can use the bilateral upsampling techniques easily. We also use the fast version of bilateral filter to improve the performance. Compared with traditional upsampling methods, our method can keep the salient structures as well as producing smooth upsampled texture results, while the computational and memory costs are also low.