Area Estimation with Subpixel Accuracy for Industrial Imaging Systems

Industrial imaging systems used for product quality control often need to make an accurate estimation of an object's area. Usually the area is estimated by counting the number of pixels classied as belonging to the object. However, with a technique called decomposition, the number of pixels can be estimated more accurately at the same resolution, or as accurately at a lower resolution. Another way to reduce the hardware costs is to lower the spatial sampling rate. The negative eeects on the area estimation accuracy can be counteracted by reconstructing the original image using interpolation. In this paper, we present the results of two experiments using HP ScanJet 4c images in which the capabilities and limitations of decomposition and interpolation were assessed quantitatively. It appeared that the resolution could be lowered with a factor 25 without decreasing the accuracy of the estimation. A reduction of the sampling rate, on the other hand, could not be compensated for as well. However, interpolation did seem to restrict the decrease in accuracy somewhat, although the results were not entirely conclusive due to hardware related reasons.