Fast target extraction from thermal imagery

1. A method for extracting a target from a series of images comprising the steps of:

(a) estimating an ambient temperature value of pixels in the series of images;

(b) finding a band of pixel values having temperature values above the ambient temperature value, the band of pixel values forming a histogram;

(c) differentiating the histogram to estimate a threshold;

(d) extracting the target having pixel values above the threshold; and

(e) colorizing the target for display;

wherein step (c) includes

computing a first derivative of the histogram,

computing a second derivative of the histogram; and

forming the threshold based on boundaries set for the first and second derivatives.

2. The method of claim 1 wherein step (a) includes

forming a first subset of pixel values of a current image defining a window of pixel values having a predetermined temperature variation; and

computing the ambient temperature value of the first subset of pixel values of the current image.

3. The method of claim 2 including the step of:

filtering the computed ambient temperature value, based on a previously computed ambient temperature value of a subset of pixel values belonging to a previous image.

4. The method of claim 1 wherein step (b) includes

finding a minimum threshold value above the ambient temperature value; and

selecting a second subset of pixel values of a current image having values above the minimum threshold value, the second subset of pixel values forming the histogram.

5. The method of claim 1 including the step of:

filtering the first threshold, based on a previously formed threshold belonging to a previous image.

6. A method for extracting a target from a series of images comprising the steps of:

(a) estimating an ambient temperature value of pixels in the series of images;

(b) forming a band of pixel values of a current image, the band comprising an amount of pixels less than an amount of pixels in an image;

(c) clustering the band of pixel values into a target class and a background class;

(d) extracting the target having pixel values above the new threshold value; and

(e) colorizing the target for display;

wherein step (c) includes

setting an initial threshold value in the band of pixel values for separating the target class from the background class, and

computing a mean value of the target class and a mean value of the background class, and

setting a new threshold value based on the two computed mean values.

7. The method of claim 6 wherein step (a) includes

forming a first subset of pixel values of a current image defining a window of pixel values having a predetermined temperature variation; and

computing the ambient temperature value of the first subset of pixel values of the current image.

8. The method of claim 7 wherein step (a) includes

filtering the estimated ambient temperature value, based on a previously estimated ambient temperature value of a subset of pixel values belonging to a previous image.

9. The method of claim 6 wherein step (b) includes

forming the band of pixel values ranging below the ambient temperature value by a first amount and above the ambient temperature value by a second amount,

wherein the first amount is different from the second amount.

10. The method of claim 6 wherein step (d) includes

estimating a distribution value of the target class, and

if the distribution value is above a predetermined value, then the target is determined to exist.

11. The method of claim 6 wherein step (d) includes

filtering the threshold value of a current image based on a threshold value of a previous image.

12. A method for extracting a target from a series of images comprising the steps of:

(a) estimating an ambient temperature value of pixels in the series of images;

(b) finding a first band of pixel values of a current image having temperature values above the ambient temperature value, the first band of pixel values forming a histogram;

(c) smoothing the histogram to estimate a first threshold;

(d) forming a second band of pixel values of the current image;

(e) clustering the second band of pixel values into a target class and a background class;

(f) computing a second threshold value between the target class and the background class;

(g) extracting the target based on the first and second thresholds; and

(h) colorizing the target for display;

wherein step (c) includes

computing a first derivative of the histogram;

computing a second derivative of the histogram; and

forming the first threshold based on boundaries set for the first and second derivatives.

13. The method of claim 12 wherein step (g) includes

combining the first and second thresholds, using a weighted average, to form a third threshold, and

extracting the target using the third threshold.

14. The method of claim 12 wherein step (e) includes

setting an initial threshold value in the second band of pixel values for separating the target class from the background class; and

computing a mean value of the target class and a mean value of the background class; and

step (f) includes

setting the second threshold value based on the computed mean values.

15. The method of claim 12 wherein step (a) includes

forming a subset of pixel values of a current image defining a window of pixel values having a predetermined temperature variation; and

computing the ambient temperature value of the subset of pixel values of the current image.

16. The method of claim 12 including the steps of:

(i) filtering the first and second threshold values of a current image based on first and second threshold values, respectively, of a previous image;

(j) forming a first subset of pixel values of a current image defining a window of pixel values having a predetermined temperature variation; and

(k) computing the ambient temperature value of the first subset of pixel values of the current image.