A new image enhancement algorithm with applications to forestry stand mappingThe theory and applications are presented of a new image enhancement algorithm which refines computer classification maps of multispectral data. The refinement eliminates connected sets smaller than a prespecified size and merges them to the surrounding area. Conventional practices in forestry timber stand mapping requires small geographic areas to be absorbed by surrounding large areas to form homogeneous stands. This homogeneity is often incompatible with the statistical formulation of homogeneity. Elements within a timber stand which should be labeled as one feature often correspond to more than one class mapped by existing computer classification techniques. The new algorithm is designed to postprocess classification maps to result in more usable timber stand maps. The new image enhancement technique is compared with an accepted neighbor-checking postprocessing technique, demonstrating the superiority of the new technique for forestry stand mapping.
Document ID
19760087397
Acquisition Source
Legacy CDMS
Document Type
Conference Proceedings
Authors
Kan, E. P.
Lo, J. K. (Lockheed Electronics Co., Inc. Houston, Tex., United States)
Smelser, R. L. (U.S. Department of Agriculture, Forest Service, Houston Tex., United States)