Abstract
The decomposition of images with great size using the Bidimensional Empirical Mode Decomposition (BEMD) necessitates an important calculation time. To overcome this problem we first proposed in [1] a new approach using Block-based BEMD method (BBEMD) where an input image is subdivided into four subblocks, and then the BEMD is applied on each of the four blocks separately. This method offered a good solution to the calculation time, unfortunately blocking artefact is noticeable in the resulting modes, this is a result of ignoring the interblock correlation during the BEMD process and every block was taken as an independent entity when interpolating the blocks borders in the sifting process. The Lapped Block-based BEMD (LBBEMD) is then proposed to whelm the artefact problem using for every block the neighbouring information by enlarging the block neighbourhood related to the two adjacent blocks inside the image during the blocks decomposition. Enlarging neighbourhood blocks necessitate intensive simulations to have for each block the corresponding enlargement size, thing that affects remarkably on the calculation time. Experimentation shows that the LBBEMD can not handle correctly the artefact problems, because the enlargement size can not be enough "or too big" to surround all the concerned pixels for each block in the image. The problem with the LBBEMD is to exactly define, a priori, for each block the size of the neighbouring enlargement size. In this paper, we will present an adaptive and fast Block Empirical mode decomposition to eliminate definitively the artefacts problems when subdividing an image into lapped blocks and we will prove, by simulations, that our method offers a good trade-off between computational time and decomposition quality.