Abstract: Image transformation and selective inverse transformation is implemented by performing DWT-based transformation on very large images using single or multi-CPU architectures without requiring large amounts of computer memory.
The compression of a large image I(x,y) is accomplished by defining L=(log2(x)−log2(Filter_size)) DWT levels, each level Ln containing pre-allocated memory buffers sufficient to hold (Filter_size+1) lines of x/2n pixels length of DWT subbands. For each line of I(x,y), in a recursive fashion through levels L0 to Ln, the level DWT is computed and the level subband wavelets are compressed and stored or transmitted as required. The recursive DWT for each level of each line of I(x,y) results in a seamless DWT for I(x,y).