This research investigates the automatic dimensional check of symmetrical parts thorough radiography and image processing. Automating dimensional checking process needs automatic feeding of parts which is achieved using vibratory bowl feeder and designing suitable traps thorough feeder’s part track that orients the parts into desired orientation.Part releasing, oriented parts transmission to x-ray imaging section and measured part’s escapement mechanisms have been integrated and designed. Designed Separation mechanism, separates measured parts into three batches: accepted, rejected and modifiable. Reducing cost, efficiency and simplicity has been considered in design process so that the full automation has been achieved with only two electromagnet and four pneumatic jacks.Monte Carlo based software MCNPX is used to simulate radiographic images to study the effects of various factors on the radiographic image. To confirm the performance of simulation software, radiographic images have been taken by a digital mammography system. Comparing the taken images with simulated ones show a good agreement. By considering penumbra effect on reducing image quality, minimum requirements to reach the 0.01 mm precision in dimensional check have been introduced. Laplacian of a tow dimensional Gaussian distribution is used to define Image edge detection algorithm. Slope value of zero crossing in second derivative histogram of each row of the image is defined as a criterion for edge detection. Effect of variables such as source voltage, part’s axis distance to the source’s focal spot and secondary filters on images is simulated and for the first time in this study, optimum value of each factor was obtained. Algorithms based on finding edge position and subtracting taken image by stored image derived from obtained dimensions is defined to detect structural and geometrical (concentricity and elliptical surface) defects respectively. Edge detection algorithm has been implemented in industrial software Labview and its performance on taken images has been investigated. Despite 422 × 528 pixel image resolution 0.03 mm precision is reached comparing calculated values derived from the algorithm with real ones. It is shown that accessing the research objective measurement precision 0.01 mm is possible through using appropriate X-ray generator and the detector. Output precision of image processing algorithm for computing internal and external angle and diameters is reached to 0.1 degree and 0.005 mm respectively from simulated images with 4096 × 3328 pixel detector resolution. This accuracy would cover a wide range of industrial parts which makes automatic dimensional check of symmetrical parts, using radiography and image processing. Key words: Dimensional check, automation, radiography, image simulation, MCNPX, image processing