FR2696260A1 - Removal of line noise from optronic images - uses vertical neighbourhood averaging as first filter then generates moving average which is subtracted from raw image - Google Patents

Abstract

The method involves passing the image data through a vertical high pass filter with limitation of amplitude to a set threshold. The filter output is then subject (4) to moving averaging and the filtered moving average is subtracted from the raw image. The high pass filter compares the current point with points in the corresponding vertical line of the image and subtracts the average luminance in this vertical group from that of the point. The subtractor is coupled to a memory holding the raw image and to the moving average calculator. ADVANTAGE - Removes bias from line noise filters used in optronic imaging devices generating scanned image.

Description

Method for delimiting optronic images
and device for its implementation
The invention relates to the field of optronic images obtained for example by infrared scanning sensors or by optronic monitoring systems, and more particularly to a process for delimiting optronic images and a device for its implementation.

 The "FLIR" type of sensors, or the panoramic watch systems making it possible to obtain optronic images present a defect that is classically called "lineage noise". To solve this problem, current systems perform averaging on each line but, given that the lineage is not completely parallel to the line, and this is likely related to the structure of the sensors (optics, sensor) and Associated electronics faults, a correction function of the whole line introduces a local bias and does not allow a sufficient correction.

 The subject of the invention is a process for delimiting optronic images making it possible to avoid the drawbacks mentioned above.

According to the invention, a process for delimiting optronic images is characterized in that it consists of starting from a noisy source image to be made
in a first step, vertical high pass filtering with amplitude limitation at a given threshold,
in a second step to apply to the filtered image thus obtained an averaging sliding in line,
in a third step to subtract this filtered and averaged image from the source image, the resulting image being devoid of lineage noise.

 The invention also relates to a device for implementing the method.

 The invention will be better understood and other features will become apparent from the description which follows with reference to the single appended figure. This figure is the block diagram of the edger device according to the invention.

 The device comprises a source image memory 1 connected to a delimited output image memory 2 via a chain comprising a three-input adder-subtractor and an output.

On the first input is applied the signal from the source image memory and characteristic of the optronic image from the sensor. The second input receives a value characterizing lineage noise that is subtracted from the current value of the source image. This noise value is obtained from the image signal as follows. The output of the source image memory 1 is connected to the input of a vertical high pass filter comprising a limiter. This device has the following function: vertical high-pass filtering makes it possible to detect the lineage noise and to extract it, the limitation introduced by the limiter taking into account the fact that the lineage is not greater than a certain predetermined level. above this level are information relating to the horizontal contour and therefore to the information itself, this information obviously having to be preserved.

In the example of the single figure, the high pass filter 3 performs an averaged "moy" on a given number of 2n + 1 points including the current point noted "c" of the image and for example the n = 4 points located in the vertical direction of the image on both sides and on the same column as the current point following the relationship
i = + 4
mean = X, im ~ s [l + i] [c] / 9 (1)
i = -4 in which im ~ s [l + ij [cJ denotes an image point at the intersection of the column "c" and the line "l + i".

The filter 3 then subtracts the average "moy" obtained from the luminance value of the current point of the source image noted im ~ s [lj [c] to which is added a constant value equal for example to 128 for normalized values of luminance staggered from 0 to 255. The value ima ~ phv [lj [cj which is thus obtained is then bounded according to the relation
128-max amp <ima # phv [lj [cj <128 + max ~ amp where max ~ amp represents the maximum amplitude of noise due to lineage in the noisy original image.

 The output of the vertical high pass filter with limiter 3 is connected to the input of a sliding averaging device operating from a number p of points of this line. This averaging only on part of the line avoids extracting dynamic noise information characteristic of vertical contours. It is this average noise value available at the output of the averaging device 4 which is subtracted from the image signal coming from the source image memory 1. As illustrated in the single figure, the third input of the adder- subtractor receives a fixed value, 128 in the example shown, corresponding to the average level of the image around which will be established variations detected and stored in the source image memory after the noise has been extracted as explained ci -above. The resulting signal from this adder-subtractor 5 is the input signal of the delayed image memory 2.

 The deliming device described above eliminates lineage noise present in infrared images from FLIRS sensors or panoramic watch systems.

The processing method implemented in this device can be summarized as follows: from a noisy source image, the method consists of
to carry out vertical high pass filtering with amplitude limitation,
- then to apply to this filtered image an averaging sliding in line,
and finally subtract this filtered and averaged image from the source image.

 This method is simple and relatively easy to implement, moreover it does not introduce an undesirable effect on the resulting image.

Claims (3)

 1. A process for delimiting optronic images, characterized in that it consists of starting from a noisy source image to be made:  in a first step (3) vertical high pass filtering with amplitude limitation at a given threshold,  in a second step (4) to be applied to the filtered image thus obtained an averaging sliding in line,  in a third step (5) to subtract this filtered and averaged image from the source image, the resulting image being devoid of lineage noise.  2. Method according to claim 1, characterized in that the high pass filtering step (3) consists of averaging "aver" over a determined number of points comparing the current point and n points located in the vertical direction of the image on the same column and on both sides of the current point, and to subtract the average obtained from the luminance value of the current point of the source image.  3. Device for implementing the method according to any one of claims 1 and 2 characterized in that it comprises a subtractor circuit (3) coupled to a source image memory I directly by a first input and indirectly on a second input via a vertical high pass filter with limiter (3) through an in-line averaging device (4).