Computer vision based on digital image processing is a fast emerging as a research tool in various branches of engineering. In manufacturing engineering environment it is mainly used for robot applications like inspection, recognition and navigation. In this work the area of application is extended to incorporate its use in metrological applications where the objective of the work had been to evaluate the surface roughness of cylindrical machined surface (turned) using machine vision technique. Basically to estimate the roughness of a cylindrical surface based on the images would pose a fundamental problem as the surface being a doubly curved one. This image distortion need to be resolved before proceeding to the evaluation of roughness of such surfaces. This study uses involved the geometric correction technique by developing an algorithm in which distortion encountered in the projection of cylindrical machined surface image is rectified. The quantification for surface roughness after opening the surfaces is performed using the surface image parameters (spatial frequency (F), arithmetic mean value (Ga) and standard deviation (STD)). Then the Group Method of Data Handling (GMDH) technique was used to obtain an analytical relationship of the roughness parameters calculated using the digital surface image and the stylus instrument values. An analysis based on the comparison to make sure that the present approach of estimation of surface finish based on the digital processed image could be implemented in practice, is presented in this paper. Copyright© (2006) by the International Measurement Federation (IMEKO).