next up previous contents index
Next: cml_draw Up: Reference Previous: cll_remove   Contents   Index

cml_decompose

$ \bigcirc$Name


cml_decompose Compute all the cmorpho_lines of a color image




$ \bigcirc$Command Synopsis


cml_decompose [-c cmimage_in] [-o ml_opt] [-l L] image_in cmimage



-c cmimage_in : original image in Cmimage structure

-o ml_opt : choose form of morpho_lines

-l L : Minimal level lines length to be kept

image_in : original color image

cmimage : cmimage with all morpho_lines




$ \bigcirc$Function Summary


Cmimage cml_decompose (cmimage_in , ml_opt , L , image_in )

Cmimage cmimage_in ;

int *ml_opt ;

int *L ;

Cfimage image_in ;




$ \bigcirc$Description


This module computes a decomposition of image_in into cmorpho_lines. It works as the module ml_decompose, but if the later assumes gray level images only, cml_decompose can be applied on 3-planes color images.

Please read before the description of ml_decompose, and notice that the only binary relation used on image's values to build the decomposition is the total order $ \leq$ defined on reals (an ordering relation is total if two elements are always comparable) . Thus, the extension of the algorithm to multi-valued images is straightforward as soon as one can recover a total order $ \preceq$ in IR3 that fits the visual perception of geometrical structures. In [CF00], a lexicographic ordering relation is proposed in the framework of topographic maps of color images :

Let L, H and S be the values of the image in the HSI color model (L stands for Luminance or Intensity, H for Hue and S for Saturation). One sets

U1 = (L1, H1, S1) $\displaystyle \preceq$ U2 = (L2, H2, S2)
  $\displaystyle \Updownarrow$
(L1 < L2)  or (L1 = L2  and H1 < H2)
or (L1 = L2  and H1 = H2  and S1 $\displaystyle \leq$ S2).
(10)
This model tries to fit the visual perception of geometrical structures : to detect shapes, human eyes are first sensitive to luminance, then to hue and at last to saturation.

The module cml_decompose allows to compute such a topographic map of the input color image. Since the algorithm assumes the HSI color model, the planes of the input image must be HSI. Use the module cfchgchannel to convert a RGB image to a HSI one. Before calling cml_decompose, you may also want to quantize the HSI channels using cfquant so that the amount of data would be reduced.

The option -l is very useful to reduce the number of level lines in the topographic map without removing too many visual information : it fixes the minimal length a level line must have in order to be kept (by default all level lines are recorded). This operation, very similar to the grain filter , may also be done afterward by cll_remove. See also the example of processing in the documentation of cml_reconstruct.




$ \bigcirc$Version 1.0


Last Modification date : Thu Apr 15 08:22:05 2004


$ \bigcirc$Author


Jacques Froment, Georges Koepfler






next up previous contents index
Next: cml_draw Up: Reference Previous: cll_remove   Contents   Index
mw 2004-05-05