Generation of Octrees from Raster Scan with Reduced Information Loss

Having addressed the issue ofgenerating octrees from raster scan efficiently using parallel computation, it is worthwhile to ensure that the encoding is accomplished without much loss of information. Knowing that each cube usually stores an average ofthe intensity value in the cubical subspace, the encoding could lead to a "blocky" (as against ''patchy'') solid during regeneration. Retaining the entire intensity distribution in three dimensions may defeat the whole purpose of encoding. The present Chapter addresses the technique for octree encoding with minimum information loss. After considering the techniques based on vector quantization [183Gary84] [152Nasrabadi88] an efficient and easy to implement approach based on wavelet transform r84Gr~ss96] is introduced. Often the reconstructed data from octrees results in a blocking effect because of the average density information assumed in the cubical space represented by the cubes defining the solid. The average density along with the variance in the cubical space also does not improve the situation. Recently, Wavelet transforms have been shown to be efficient in giving a compressed representation of square and cubical region [1 84Gross96] and the inverse wavelet transform, reconstructs the cubical space without much loss of information. The cubical region, which is represented by an octree leaf node along with the wavelet transform of the cubical region, can be thus