In normal industrial radiography-on-film two sources of unsharpness, geometric unsharpness Ug and inherent unsharpness U_i, should be considered. However, the others can be neglected. So, U_total = (U_g²+U_i²)^1/2

If a measure is taken to reduce the total unsharpness, the density of the image will be reduced, and this will lead to the reduction of the contrast, which separates regions-of-interest from the background in the film. Geometric unsharpness, generated due to the focal size, can be reduced by increasing source-to-film distance, but that lowers the energy reaching to the work piece and so the length of exposure-time will be lengthened, and is economically undesirable. The inherent unsharpness is caused by the effects of secondary electrons generated in the film emulsion. The magnitude of the inherent unsharpness depends on the X-ray energy. Using the soft X-ray to penetrate the work piece, inherent unsharpness will be reduced, however, the density of the image will be reduced, thus the contrast of the defect with its background will be reduced. To overcome this drawback, the length of exposure-time must be increased.

Using digital image processing technique, the contrast can easily be enhanced but the unsharpness is difficult to reduce. A special radiography technique can be adopted by using the soft X- ray produced at low kilo-volt and the long focal distance.

A series of special and normal industrial X-ray photographs were taken respectively. The normal ones were taken at a source-to- film distance 600 mm and 140 kV and the special ones at source- to-film distance 600-2400 mm and 50-140 kV. The exposure-times are equal for all radiographs. The normal ones are suitable according to ASME E-142 and ISO 5579-85, but the special ones are not because of its low density, 0.8-1.5.

Considering some noise is multiplied and some is additive, image processing consists of first averaging 16 images to reduce the additive noise, then homomorphic filtering the image to reduce the multiplied noise, finally enhancing contrast.

Low-kV X-ray and long source-to-film distance radiography, according to U_g=U_i, has been developed. Film density is about 1.0 -1.2 that can ensure the high signal-to-noise-ratio which is required by digital image processing to detect very small defects, some of them can not be detected by image processing of the normal industrial radiography film.