Method of detection of points of interest in a digital image

1. A method of detection of points of interest in a digital image of a sequence of images ( 12 ) of a scene picked up by a camera ( 10 ), this method implementing an point of interest search algorithm ( 14 ) that is parameterizable with a detection threshold (τ) such that the number (N) of points of interest detected in the image varies as a function of the threshold level,

this method being characterized by a modeling by a decreasing exponential function of the characteristic giving the number (N) of detected points of interest as a function of the threshold (τ), this exponential function being dynamically parameterizable with values (C, σ) linked to an image to be analysed,

and in that it comprises the following steps:

a) determining ( 18 ), for a current image, values of parameterization (C, σ) of the decreasing exponential function;

b) predicting ( 18 ), for said current image, an optimum value of the detection threshold (τ) by using the modelled characteristic, parameterized with the values determined at step a); and

c) applying ( 14 ), for at least one image following said current image, the point of interest search algorithm with the optimum detection threshold value (τ) computed at step b).

2. The method of claim 1 , wherein said optimum detection threshold value (τ) predicted at step b) is a value corresponding to a given number (N) of points of interest, as indicated by said characteristic giving the number of points of interest detected as a function of the threshold.

3. The method of claim 1 , wherein said decreasing exponential function is a square root decreasing exponential function:

N

⁡

(

τ

)

=

C

⁢

⁢

exp

⁢

{

-

τ

σ

}

τ being the detection threshold value, brought back to a pixel-homogeneous quantity,

N being the number of points of interest detected for a threshold τ, and

C and σ being said parameterization values linked to the current image.

4. The method of claim 1 , wherein the step a) of determining, for a current image, the parameterization values of the exponential function comprises the following sub-steps:

a1) a first search for points of interest in the current image by the search algorithm ( 14 ) with a first predetermined value (τ 1 ) of the detection threshold, giving for result a first number of points of interest (N 1 );

a2) at least one second search for points of interest in the current image by the search algorithm ( 14 ) with a second value (τ 2 ) of the detection threshold higher than the first predetermined value (τ 1 ) of the detection threshold, giving for result a second number of points of interest (N 2 ); and

a3) the determination of the parameterization values C and a based on the number of points of interest (N 1 , N 2 ) obtained at steps a1) and a2).

5. The method of claim 1 , further comprising, during the iterative execution of the steps a) and b):

the application of camera auto-exposure information as supplementary input data for the prediction of the detection threshold.

6. The method of claim 1 , further comprising:

the division of the current image ( 50 ; 54 ) into a plurality of reduced-size sub-images ( 52 ; 56 );

the execution of steps a) and b) independently for the different sub-images of the current image, with, for result, an optimum value of detection threshold peculiar to each sub-image; and

the execution of step c) with application, for each sub-image of the following image, of the point of interest search algorithm with the respective optimum detection threshold value peculiar to this sub-image.

7. The method of claim 6 , wherein the optimum detection threshold value is a value corresponding to a same predetermined number of points of interest for all the sub-images.

8. The method of claim 1 , further comprising:

previously to step a), the production ( 30 , 32 ) of a multiresolution representation of the pyramid-of-images type, modelling the current image of the scene picked up at different successively increasing resolutions (QQQVGA, QQVGA, QVGA, VGA); and

the iterative execution of steps a) and b) for each level of the multiresolution representation ( 34 , 38 , 40 ), starting by the level of less resolution (QQQVGA), the detection threshold value determined for a given level being applied as an input data for the prediction of the detection threshold at the higher level of resolution

the iterative execution of steps a) and b) for each level of the multiresolution representation ( 34 , 38 , 40 ), starting by the level of less resolution (QQQVGA), the determined detection threshold value for a given level being applied as an input data for the prediction of the detection threshold at the higher level of resolution.

9. The method of claim 8 , further comprising:

the execution of the step c) with the application, for each level of the multiresolution representation of the following image, of the point of interest search algorithm with the respective optimum detection threshold value peculiar to this level.

10. The method of claim 8 , further comprising, during the iterative execution of steps a) and b), for each level of the multiresolution representation:

the application, as complementary input data for the prediction of the detection threshold at the higher resolution level, of the optimum thresholds and the applied thresholds corresponding to the lower levels of resolution.

11. The method of claim 8 , further comprising, during the iterative execution of steps a) and b), for each level of the multiresolution representation:

the application, as complementary input data for the prediction of the detection threshold at the higher resolution level, of camera auto-exposure information.

12. The method of claim 8 , further comprising:

the division of the images at the different levels of the pyramid of images ( 50 ; 54 ) into a plurality of reduced-size sub-images ( 52 ; 56 );

the execution of steps a) and b) independently for the different sub-images, with for result an optimum detection threshold value peculiar to each sub-image; and

the execution of the step c) with application, for each sub-image of the following image, of the point of interest search algorithm with the respective optimum detection threshold value peculiar to this sub-image.

13. The method of claim 12 , wherein the optimum detection threshold value is a value corresponding to a same predetermined number of points of interest for all the sub-images.