Industrial measurement of surface roughness is primarily stylus-based. Stylus-based surface roughness measurement has limitation of the stylus tip diameter acting as a filter on steep valleys. It is also a time consuming and an offline process. Vision-based system for surface roughness measurement has potential to emerge as a reliable online surface roughness measuring system, because of the emergence of powerful cameras and superior image processing techniques. The presented work is an attempt to create a technique for evaluation of surface roughness using vision-based image processing. First a group of two-dimensional images of turned surfaces was obtained and using MATLAB the image pixel intensity distribution parameters were calculated for each of these surface images. Mean and standard deviation were the two statistical parameters used to characterise the surface roughness using the pixel intensity of the turned surface images. Standard deviation was found to correlate better with Ra, Rda and Rdq values of the surface roughness. Hence, the technique may be preferred for online surface characterisation of turned surfaces. Copyright © 2017 Inderscience Enterprises Ltd.