This paper deals with the inspection of highly reflective chrome-coated rings used in textile machinery using machine vision. These rings are mass produced in very large numbers, and the inspection was done manually using an optical microscope. Introduction of vision inspection using algorithms supplied by a commercial vendor had not helped the industry to achieve 100% quality inspection. In order to improve inspection speed and to ensure 100% quality inspection, it was absolutely essential to improve the complete inspection process, and it was also required to classify defective and non-defective components by a proper sorting algorithm. The effect of the curved, reflective nature of material and the real-time inspection make the imaging and defect detection and classification difficult. In the present study, four different algorithms based on Fourier filtering, auto-median, image convolution, and single-step thresholding approaches were used for defect detection, and then their performances were compared with reference to efficiency of defect classification and speed. The complete procedure, analysis, and the results of different image processing algorithms used in defect detection are reported in this paper. © 2010 Springer-Verlag London Limited.