An image processing system was used to evaluate the number and sizes of voids formed in woven, knitted and braided fabrics having different attributes, such as their geometrical constructions and yarn arrangements. The box-counting method by computer-aided image analysis was adapted to measure the fractal dimension of their void patterns. Gray scale image analysis was also applied for the characterization of their visual texture. The angular second moment, correlation, contrast and entropy extracted from the gray level co-occurrence matrix were measured as textural feature parameters. As a result of the image analysis, it was found that the lacunarity of the knitted fabric is the most irregular and complex, due to its looped construction, and that all three fabrics had visual anisotropy as texture features. From the viewpoint of fabric construction, visual features of the three fabrics were discussed in terms of uniformity, directionality, local variation and information quantity.