Pilling segmentation for objective pilling evaluation

A pilling segmentation method was introduced for objective pilling evaluation. Traditionally, image data collected are photo-sensitive which suffered seriously from the light reflectivity from surface and color patterns on fabrics. The data used in the article were surface profile in gray level format. The background approximation by wavelet analysis approach was introduced. The original image was orthogonally decomposed and represented by sets of wavelet coefficients. The approximation coefficients were extracted at various levels and reconstructed to form a background. The results demonstrate a very effective segmentation in all tested fabrics.