AN ADAPTIVE IMAGE PREPROCESSING ALGORITHM FOR DEFECT DETECTION IN DOUGLAS-FIR VENEER

Analysis of computer images of wood can usually be enhanced by the ear ly elimination of uninteresting areas in the images. SISM and NSISM ar e algorithms that identify clear regions in images of wood so that suc h regions can be largely ignored in subsequent processing. Both perfor m well on most images, but falter on some. This paper describes how NS ISM was refined to make it perform better on troublesome image areas w ithout compromising its performance on more typical areas. Making such refinements to any scanning algorithm is a difficult undertaking that is usually characterized by modest performance gains. However, the ec onomic viability of a scanning system may depend critically upon wheth er such gains can be achieved. A comparison of the refined algorithm, called ASISM, to the original two shows that ASISM performs better tha n or equal to its predecessors across a range of commonly occurring wo od features in Douglas-fir veneer.