Fractal dimension is a commonly used method to measure texture coarseness. However, the definition of fractal dimension is very difficult to resolve. Many fractal dimension algorithms of image are proposed based on different application backgrounds. But they lack comparison and evaluation of the overall and are only limited to the selection of the box-counting dimension method. This paper makes a testing among these methods in sensitivity and time complexity for texture coarseness of image. Through a comparison and analysis to applicability of these methods, the recommended models are given at last.