This paper presents a novel approach of visual inspection for texture surface defects. The approach uses artificial immune theory in learning the detection of texture defects. In this paper, texture defects are regards as non-self, and normal textures are regarded as self. Defect filters and segmentation thresholds used for defect detection are regarded as antibodies. The clonal selection algorithm stemmed from the natural immune system is employed to learn antibodies. Experimental results on textile image inspection are presented to illustrate the merit and feasibility of the proposed method.