Surface quality of finished parts, is one the most considerable factors in the field of quality control. Surface roughness, however, is the major parameter among various surface parameters which affects directly on friction, fatigue, erosion and corrosion forces and final appearance of the surface. The traditional mean for measurement of surface roughness is the stylus instrument which contacts with surface and scans the surface profile. The traditional stylus has several disadvantages. One the most important disadvantage of stylus is that the surface roughness must be checked off-line and surface roughness control during the manufacturing process is impossible. Therefore many researches have been done on non-contact methods which are able to measure surface roughness on-line. In the present study a new optical and non-contact approach for surface roughness measurement is presented. In this method several correlation surface parameters are extracted using laser light scattering and image processing technique; by mean of artificial neural networks the surface roughness is estimated with desirable accuracy. Despite of extreme simplicity setup and requirement of low cost equipments, this method has shown reasonable accuracy which is comparable with traditional stylus instrument. In addition, because this method is non-contact, it can be applied on-line for continuous production lines. In the presented approach, laser light is radiated to various surfaces with various roughnesses and its reflection is sensed by a CCD camera. The reflection differs from incident laser light, which is due to surface roughness. By using image processing technique, several correlated factors are extracted from laser reflection images. Artificial neural networks are used to estimate the relationship between surface roughness and extracted parameters. By having adequate samples, an appropriate ANN can be trained; the trained network can estimate surface roughness of new samples properly accurate. The results of this method and stylus instrument have been presented to be compared. It can be seen that by having suitable imaging instruments, this approach can surpass traditional stylus technique. Having no contact with surface roughness, full automation feasibility, quick results and on-line production control are other advantages of this method. Keywords: Surface Roughness, Laser, Optical Measurement, Image, Artificial Neural Network, Back Propagation