The surfaces produced by electrical discharge machine (EDM) were illuminated using a redline laser source and the speckle line images of the specimens were captured using a CMOS camera. Signal vectors were generated from the speckle line images which are 1 × 2592 matrices, represented the grey scale intensity of the speckle line images. The image signal vectors were then decomposed using a wavelet transform by key local intensity variation method. The fourth and the fifth levels of the six-level decomposition of bi-orthogonal wavelet were expected to bear the details about the surface roughness. The root mean square (RMS) and variance of the 4th and the 5th level decomposition were obtained and were compared with the surface roughness parameters roughness average (Ra), arithmetic mean slope (Rda), and root mean square slope (Rdq). RMS and variance of the 4th level decomposed signals were found to correlate well with surface roughness parameters. © 2019 Elsevier Ltd