Compression by image empirical mode decomposition

Empirical Mode Decomposition (EMD) in two dimensions provides a tool for image processing by its special ability to locally separate spatial frequencies. The tendency is that the Intrinsic Mode Functions (IMFs) other than the first are low frequency images. Variable sampling of the EMD is used for image compression. This is done block-wise using the non-uniformly located extrema points of the IMF to steer the uniform sampling rate of the block. The variable sampling process results in a number of samples for each block. These can be squeezed into blocks of smaller size and DCT coded. The DCT components are then quantized and thresholded, leaving us with even fewer components to represent the block. The coding method presented in this paper is a new approach to image compression that opens doors for future applications.