EP2777271A2

EP2777271A2 - Method for carrying out a dynamic range compression in traffic photography

Info

Publication number: EP2777271A2
Application number: EP12794852.9A
Authority: EP
Inventors: Ralf Kachant
Original assignee: Jenoptik Robot GmbH
Current assignee: Jenoptik Robot GmbH
Priority date: 2011-11-11
Filing date: 2012-11-10
Publication date: 2014-09-17
Also published as: CN103931198A; CN103931198B; DE102011055269A1; US9153038B2; US20140355875A1; WO2013068006A2; DE112012004684A5; WO2013068006A3; AU2012334490A1; AU2012334490B2

Abstract

The invention relates to a method for carrying out a dynamic range compression in traffic photography for representation having greater detail fidelity in images created in connection with traffic monitoring installations. The problem addressed by the invention is that of finding a possibility for achieving, in the case of digitally obtained images in traffic photography, whilst precluding the different subjective influences on the part of the processing personnel, a representation of the dark regions with greater detail fidelity, without the information of the brighter regions being lost in the process. According to the invention, this problem is solved by means of a method for carrying out a specific dynamic range compression in traffic photography.

Description

Method for performing dynamic compression in traffic photography

The invention relates to a method for performing dynamic compression in traffic photography for more detailed representation in images that were created in connection with traffic surveillance systems.

In traffic photography mostly pictures are created in which the driver in the vehicle interior very dark, on the other hand, the indicator appears greatly outshined by its retroreflective property. In order to lighten the dark areas in high dynamic range images and to make low contrast details in those areas more visible without destroying the information in the brighter areas, dynamic compression can often be successfully used. Here, a copy of the image is converted into a contrast mask and overlaid with the image. For each image, brightness and contrast must be adjusted manually, so that the corrected images do not look unnatural or even like a photomontage.

If now the images of the traffic monitoring systems in the evaluation offices (back office) of the authorities or commissioned organizations are evaluated, daily several thousand pictures must be adjusted manually, which brings a considerable temporal overhead, but also an additional load of the personnel with itself. In extreme cases, there may be a limitation of fines if there is a congestion in the processing of the images. Furthermore, the manual adjustment is subject to the subjective influences of the individual who processes the image.

The object of the invention is to find a way to achieve a more detailed representation of the dark areas in digitally obtained images of traffic photography, excluding the different subjective influences from the editing staff, without losing the information of the brighter areas. According to the invention, this object is achieved by a method for carrying out a dynamic compression in traffic photography. Starting from a digitally present image original with the pixels x _F i, divided into n columns s and m rows z:

First, a gray scale image mask [G] with the gray values G (Sii; z _i2 ) is created, where = {1, ..., m} and i ₂ = {1, ..., n}. In the further course of the procedure, first the arithmetic mean value is calculated over all gray values G (Sii; z _i2 ) according to

Λ m n

9 = * Σ TG (s _n ; z _i2 )

n * ^m i1 = 1 i2 = 1 determined or calculated. Since the mean value g is merely an arithmetic mean value x, all the pixels XQ of the gray value image mask with i = {1,..., N * m} can also be summed up and calculated according to the following formula: n * m

^; I (x _G i) - n * mi = 1

From the arithmetic mean value x, a gain parameter p and a blur factor b are then determined. Subsequently, a blurred gray value image (contrast mask) [U] with the gray values Xu, by blurring with the Blur factor b is created to finally the desired new image [N] with the pixels x _N i by superimposing the image original [F] with the pixels x _F i with the blurred gray value image (contrast mask) [U] with the pixels Xu, with the inclusion of Gain parameter p and an exponent Vu.

The particular advantage of the method according to the invention is that the dark areas are automatically lightened, particularly in the case of color images with high dynamics, in order to better visualize details that are initially not visible in these areas, but without destroying the information in the brighter areas. In traffic photography, this is particularly advantageous because the driver is usually shown very dark, while the license plate is greatly outshined with its retroreflective properties. This method can be applied both in the camera immediately after the detection in such a way that all the image original are automatically lightened or after the image sources have been stored on a storage medium. The temporally downstream brightening can take place in the camera at a time in which no image original is detected, or on a separate arithmetic unit, for example on a computer in a back office.

The starting point for the method can be both a monochrome gray value image [F] and a color image [F] with the usual 3 channels (RGB). In the case of the color images, the gray value image mask with its gray values x _G i is thereby generated by converting the individual R, G and B pixels as follows:

When finally generating the new color image [N], each pixel is then split again into an R, G and B pixel.

To largely suppress rounding errors, it is advantageous to convert the calculator, which is usually set to fixed-point arithmetic, to floating-point arithmetic. The invention will be explained in more detail with reference to embodiments. In the accompanying drawings show:

Fig. 1 shows a schematic sequence of the method based on a color image and

2 shows a schematic sequence of the method based on a monochromatic Bildoriginals.

Since all computers today use so-called floating-point, floating-point, floating-point or floating-point arithmetic ("floating point"), at the beginning of the process all color values from fixed-point arithmetic (integer) to floating-point arithmetic, i. H. Floating point numbers converted to keep rounding errors in the calculation steps as low as possible.

For conversion to a floating-point number, each R, G and B value for each pixel is divided by 65,536.0:

Int -> floating point

65,536.0

In the first step of a digitally present image original [F] with the pixels x _F j with i = 1, ..., n * m, first a gray scale image mask [G] with the

Gray values x _G i, where i = 1, ..., n * m, created. The starting point for the method can be both a monochrome gray value image [F] and a color image [F] with the usual 3 channels (RGB).

For a color image [F], there is an R, -, G, - and Bi value for each pixel x _Fj . The Rr, Gi and Bi values represent a value triple for each pixel. On the other hand, for a monochrome image original, there is only a single gray value for each xpj. In the case of a monochrome image original [F], only one copy of the image original [F] is generated, which then corresponds to the gray scale image mask [G].

In contrast, in the case of a colored image original, the color image [F] is first converted into a monochrome image. For this purpose, a copy of the image original [F] is first generated from the color image [F] and subsequently converted into a monochrome image [G]. For this purpose, the gray value image [G] with its gray values x _G i is generated by converting the individual R, G and B pixels as follows:

R; + G; + B:

Gray values XQ = - ^{■ L} ,

3 so that the calculated brightness values are then assigned as gray values to the corresponding pixels x _G i of the gray value image [G] to be generated.

In the second step, the gray value image [G], the arithmetic mean x over all gray values XGI, with i = 1, ..., n * m, calculated from a monochrome image original [F] or from a color image [F ] was generated, calculated or calculated:

1 n * m

Arithmetic mean x = * £ (x _Gi ).

n * m; _ Λ

In the third step, using a blur filter, a blurred gray scale image is generated for later use. The blur filter (eg a Gaussian filter) reduces the difference between the gray values (contrast) between adjacent points. Since it is a low-pass filtering, small structures are lost, but large ones are preserved. A common blur filter is a Gaussian filter in which the gray value of each pixel of the contrast mask x _U i for a two-dimensional image is defined by x _ui = h (x, y) by the formula: x + y

1 -

h (x, y) = ^ e ^2σ '

With reference to the method according to the invention, this means that s is to be used for x, z for y and b for σ. Here b is the uncertainty factor, which is calculated from the arithmetic mean value x of the gray value image [G] with the following formula b = 4 * x, wherein the functional relationships were determined empirically from several image series with different illumination situations. To reduce the computational effort, a two-dimensional filter H (x, y) can be separated into two one-dimensional filters H (x) and H (y). Thus, this is applied, for example, only to the direct neighbors on the one hand in the horizontal and on the other in the vertical and thus reduces the computational effort many times.

Thus, each gray value x _ui with i = 1, ..., n * m for each point of the contrast mask [U] results by applying the Gaussian filter to every point x _G i where i = 1, ..., n * m, taking into account the contrast values of the respective neighboring points, the gray value image [G].

In the fourth step, the original image [F] is compared pixel by pixel with the blurred gray value image [U] (contrast mask). The calculation takes place as a function of a gain parameter p determined once for all calculations of an image and of the pixel-dependent parameter x _ui, which both flow into the exponent specific for each pixel. This kind of calculation preserves the natural impression in the newly generated image [N]. For this purpose, gain parameter p is first used with the aid of the arithmetic mean value

10 ₊ (°°° -)

55 + x calculated.

Subsequently, the pixels R, G and B of the new color image [N] are determined for each individual pixel of R, -, G, - and Bi with i = 1, ..., n * m as follows: in the case that x _u , <0.5:

R _Ni = 1- (1-R _Fi ) ^v - G _Ni = 1- (1-G _Fi ) ^v - B _Ni = 1- (1-B _Fi ) ^v - for the case where x,

R _Ni = Rn ^Vui

G _Ni = G _Fi ^v - B _N i = B _Fi ^Vui , where in both cases the exponent is calculated pixel by pixel as follows:

(2x _ui -1)

v _{Ui =} 10l 50 where x _{ui are} the values of the blurred gray value image (contrast mask) and p is the gain parameter calculated using the arithmetic mean value x. The dark parts of the picture are amplified without changing the bright picture parts too much.

Finally, for each pixel, the flow number of the R, G, and B values are converted back to an integer integer by multiplying by 65,536.0.

Floating point -> int

-> x * 65,536.0

In FIGS. 1 and 2, the process sequences explained in greater detail above are illustrated schematically, wherein in FIG. 1 a color image and in FIG. 2 a monochromatic image are assumed.

designations

Gray value image (gray value mask) with gray values x _G i

arithmetic mean x

Gain parameter p

Blur factor b

Exponent vu,

Picture original [F]

new color image / new generated image [N]

blurred gray value image (contrast mask) [U] with gray values x, new pixels XNI

Claims

claims

1 . Method for carrying out dynamic compression in traffic photography, starting from a digitally present image original with n * m pixels, characterized by the following method steps:

- creating a gray value image mask [G] with gray values x _G i from the image original, where i = 1 n * m,

- Calculate the arithmetic mean

1 n * m

x = * E (Qi).

n * mj ₌

Determining a gain parameter p and a blur factor b,

- Creating a blurred gray value image [U] with gray values x _ui by blurring the gray values XQ with the blur factor b and

- Create a new image [N] by superimposing the image original [F] with the blurred gray scale image [U] using the gain parameter p.

2. The method according to claim 1, characterized in that when using a monochrome image original [F] the gray scale image mask created [G] is a copy of the image original.

3. The method according to claim 1, characterized in that when using an image original [F] consisting of R, G and B pixels, the gray scale image mask [G] is created by the gray values XQ are calculated by

_ R _{i +} G _{i +}

x _Gi -

Method according to one of the preceding claims, characterized in that the blur factor b is calculated by

5. The method according to any one of the preceding claims, characterized in that the gain parameter p is calculated by

10 ₊ A 55 + x

6. The method of claim 1, 2 and 5, characterized in that the new pixels XNI are determined at x _U i <0.5 by x _Ni = 1 - (1 - x _Fi ) ^v and at x _ui > 0.5 by

X _N i = Fi

(2x _ui -1

50

where Vy j = 10.

A method according to claim 1, 3 and 5, characterized in that the pixels RNI, GNI, B _N i are determined at x _U i <0.5 by and at x _ui > 0.5 by

^R Ni = ^R Fi ^Vui . ^G Ni = ^G Fi ^Vui . ^B N1 = ^B F | ^V

(2x _ui -1)

50

where Vy j = 10.