Texture synthesis using gray-level co-occurrence models: algorithms, experimental analysis, and psychophysical support

The development and evaluation of texture synthesis algorithms is discussed . We present texture synthesis algorithms based on the gray-level co-occurr ence (GLC) model of a texture field. These algorithms use a texture similar ity metric, which is shown to have high correlation with human perception o f textures. Synthesis algorithms are evaluated using extensive experimental analysis. These experiments are designed to compare various iterative algo rithms for synthesizing a random texture possessing a given set of second-o rder probabilities as characterized by a GLC model. Three texture test case s are selected to serve as the targets for the synthesis process in the exp eriments. The three texture test cases are selected so as to represent thre e different types of primitive texture: disordered, weakly ordered, and str ongly ordered. For each experiment, we judge the relative quality of the al gorithms by two criteria. First, we consider the quality of the final synth esized result in terms of the visual similarity to the target texture as we ll as a numerical measure of the error between the GLC models of the synthe sized texture and the target texture. Second, we consider the relative comp utational efficiency of an algorithm, in terms of how quickly the algorithm converges to the final result. We conclude that a multiresolution version of the "spin flip" algorithm, where an individual pixel's gray level is cha nged to the gray level that most reduces the weighted error between the ima ges second order probabilities and the target probabilities, performs the b est for all of the texture test cases considered. Finally, with the help of psychophysical experiments, we demonstrate that the results for the textur e synthesis algorithms have high correlation with the texture similarities perceived by human observers. (C) 2001 Society of Photo-Optical Instrumenta tion Engineers.