WO2011000392A1

WO2011000392A1 - Method and camera system for improving the contrast of a camera image

Info

Publication number: WO2011000392A1
Application number: PCT/EP2009/004783
Authority: WO
Inventors: Christopher Gideon Reade; Patrick Eoghan Denny
Original assignee: Hi-Key Limited; Application Solutions (Electronics And Vision) Ltd
Priority date: 2009-07-02
Filing date: 2009-07-02
Publication date: 2011-01-06

Abstract

A method and a camera system for improving the contrast of a camera image comprising an evaluation of the brightness distribution of the image and application of a contrast correction function to the image. In order to improve the contrast for difficult lighting conditions the spatial brightness distribution is determined, whereas the contrast correction function is varied depending on the spatial brightness distribution.

Description

Method and camera system for improving the contrast of

a camera image

Prior art

The invention relates to a method and a camera system for improving the contrast of a camera image, such as preferably used in driver assistance systems for motor vehicles, according to the preambles of Claim 1 and 18.

Camera images or video images are used in image- or non- image evaluating driver assistance systems, or else what are referred to as vision-based driver assistance systems, for sensing the surroundings, and in the known syscems objects, obstacles, rcαdv.'ay boundaries and lane boundaries as well as distances thereto are determined on the basis of the camera images. In order to capture the actual images, image sensor, which are referred to as imagers are used in what are referred to as vision- based surroundings-sensing systems and they then supply a camera image of the sensed surroundings, which is also referred to as a sensor image and whose pixels or image areas can be evaluated in terms of intensity, contrast, colour or other parameters using

corresponding data-processing means. EPl 475 745 Al discloses such an evaluation of pixels of a video image, in which evaluation of a frequency distribution of the brightness values over the surface of the image sensor is performed. Areas are defined in which the pixels are evaluated in order to improve the contrast and the brightness values of the pixels are changed by means of a correction data calculation by comparing the brightness values with pixels which are associated with surrounding areas . If a camera in, for example, a vehicle captures a scene on a roadway, the brightness content of the scene should be as far as possible in a range which covers a numerically appropriate range for evaluation in the computer system. If a scene has high dynamic range here, i.e. there is a large difference in terms of brightness between different areas of the scene, decision processes may be necessary in the system to permit prior treatment of the data for effective processing of the contrast in such a way that at the same time a high-resolution contrast is ensured at all brightness levels in a limited area of the numeric output data. A series of algorithms, by means of which in particular in common cameras with limited dynamic range (non-HDR cameras) a decision can be made in terms of the determine whether a range of brightness values is found which represent sufficiently the scene to be captured and in doing so still sufficiently retain the structure of the scene, are already known per se . It is essential here that a medium degree of brightness in the scene is possibly used for working on the captured video image, in which case, for example, the mean value of the numeric output data of the video image corresponds approximately to the mean range of the brightness of the scene. In the two above mentioned cameras with limited dynamic range it is frequently found that the dynamic range of the possible output data is greater in terms of the brightness than the representation capability of the camera. In this context, it is then necessary to decide how the captured video information is to be displayed in an optimum way. To ensure that the camera operates in an appropriate, medium range of the output image data, the camera usually changes the exposure values and/or the amplification values, but this entails the effect that if it reduces the amplification and/or the exposure to an excessive degree, the relatively bright areas of the image have content and structure but the relatively dark areas of the image are then virtually impossible to evaluate. If the amplification and/or the exposure are increased, the relatively dark areas of the image exhibit content and structure but the

relatively bright areas of the image are then virtually impossible to evaluate.

It is also obvious that the camera itself decides on the general exposure in the sense that it automatically adjusts the exposure of the camera to give a sensible picture. If the camera evaluates that a scene looks particularly bright it reduces the exposure time and if it looks particularly dark it increases the exposure time. There is a specific pair of routines that do this in the camera, one of which is typically called the Automatic Exposure Control (AEC) and Automatic Gain

Control (AGC) . The first of these increases the signal value from a pixel by letting in light for a longer period. The second of these controls amplification of the signal from a pixel by increasing the gain on an amplifier coming from a pixel.

If the driver of the vehicle is looking at a scene in normal daylight the picture looks good. In this case the exposure of the camera is probably well chosen to give a good picture and there is enough light so the voltages recorded in the pixels don't need to be artificially amplified by changing the amplifier gain. So the exposure time is reasonable and the gain is low. If the light level drops, it is common to increase the exposure time to let in more light. This is an

appropriate method of getting more data from the scene, because doing so does not add noise to the pixels, but increasing the gain on any amplifier will make a signal noisier, which could cause problems at low light.

The most imager AEC systems are working in the

following manner:

a) capture a frame

b) count how many "bright" pixels are in the picture c) count how many "dark" pixels are in the picture d) If the number of "bright" pixels is too high, reduce the exposure time if possible

e) If the number of "dark" pixels is too high, increase the exposure time if possible

f) go back to step a) Steps d) and e) above subject to a hysteresis to prevent the overall image changing too fast when some features in the scene become brighter or darker, otherwise it would be possible in a driver assistance system to start the driver or may cause him too much variation in the input image, which might confuse the other algorithms on the image sensor.

When the light levels are dropped, the exposure time increases until it reaches a limit. This limit is usually related to the frame rate, since it is not possible to expose an image longer than a frame,- if it's done it is not possible to output a frame within the frame period. When the exposure reaches the limit, there isn't really any dynamic AEC and the AGC starts to work, increasing the gain. Eventually, when the light levels are dropped to almost complete darkness the exposure is at a maximum, the gain is at a maximum and the picture is dark and noisy. Conversely when the light level is increased, the AGC gain drops to a minimum, the AEC drops the exposure to a minimum and eventually a completely white image is obtained. Summary of the invention

The invention is based on a method for improving the contrast of a camera image comprising the steps of evaluating the brightness distribution of the image and applying a contrast correction function to the image, whereas evaluating the brightness distribution includes determination of a spatial brightness distribution of the image, and whereas the contrast correction function is varied depending on the spatial brightness

distribution in the image.

The advantage archived with this method is that the spatial brightness distribution can be used to detect certain, a prior known lighting condition, in which an automatic gain or exposure control of the camera would lead to unsatisfying results. If such lighting

conditions have been detected the correction function is automatically varied accordingly in order to improve the contrast distribution of the image.

Such lighting conditions typically occur at low sun level or generally at dusk or in situations in which the foreground is mostly in shadow. In these cases significant areas of the camera image are relatively dark and others are relatively bright, so that the difference in brightness between the areas is

comparable or even higher than the dynamic range of the camera. In other words the image comprises a great upper area containing the bright sky and lower area containing the significantly darker foreground. These lighting conditions will be referred to as Bright Sky Darkening Foreground (BSDF) . Without the features according to the invention, the camera could therefore reduce the enhancement and/or the exposure to such an degree that the foreground would be dark and low in contrast for optimum image evaluation.

The method according to the invention is designed to detect such situations by evaluating the spatial distribution of the brightness in the image and to vary the correction function accordingly in order to improve the contrast distribution. The contrast distribution is preferably corrected such that the contrast is

increased in the dark, foreground area of the image, which usually contains the more important image

information, and decreased in the bright, sky area of the image. This means that the method according to the invention advantageously makes use of priori knowledge for both detection of certain difficult lighting conditions and decision on how the contrast

distribution is changed.

Such a priory known, difficult lighting conditions could also occur in connection with vehicle mounted cameras for driving assistance systems, if, for example the image taken by the camera contains an bright area due to reflections at the vehicles body. In particular these reflections can be reflections of light from a vehicle light at a vehicle body part. Other examples of such lighting conditions include reflections of light at a wet road surface.

The different areas can, however, also be defined by a relatively dark, inner area which is more important for the evaluation, in particular by entrances to tunnels or garages, and by a surrounding, relatively bright area. It is also possible for the different areas to be defined by a relatively bright area owing to

reflections or salient colouring, and a relatively dark area which is more important for the evaluation. In this context, the areas which are relatively bright due to reflections or salient colouring can be parts of the vehicle carrying the cameras.

According to the invention, the method for evaluating the image data is carried out with essentially two method blocks. On the one hand, heuristic detection and, on the other hand, correction of the image data is performed. In the detection phase, a structure of the captured scene is initially captured as a problematic state in the sense of a critical bright/dark distribution, and, in the most frequent case, this is, for example, what is referred to as a BSDF structure with a bright sky and a darkening foreground (BSDF = bright sky darkening foreground) . The data stream which contains the image data, i.e. the brightness or the chroma values, at the output of the carriers, is subsequently checked in the image-evaluation means to determine whether the

abovementioned criteria are met. This may involve preferably heuristic methods in which sufficient definitive information about the structure of the image can be acquired by evaluating less image data, i.e. in order to see whether areas of the data stream

correspond partially or entirely to such a structure of the image .

The upper area of the image is usually bright and relatively homogenous in this context, and the lower part is dark and also relatively homogenous. It is therefore possible for a distribution of the image data to be processed to be performed as a function of the brightness of the area in accordance with the

abovementioned BSDF structure, with the bright and dark areas being processed with different numerical criteria during the image-evaluation process. This means, in particular, that these numerical criteria are

attenuated or amplified in order to correspond to the realistic perception of the image. An important point here is that the heuristic method is configured in such a way that a high level of reliability is obtained, in which case it can be proven experimentally that even simple heuristic methods with evaluation of a small number of image data items produce good results. The detected spatial brightness distribution is compared to an predetermined brightness distribution in order to determine, if BSDF or equivalent predetermined lighting conditions are present.

The determination of the spatial brightness

distribution preferably includes determination of the brightness at predetermined image locations. In this way the a priory knowledge is used to reduce the necessary image processing amount.

The όpdLidl bx iyhLness distribution of the image can be evaluated along at least one predetermined geometrical axis or curve. The axis can be a vertical axis in order to detect a BSDF situation. For other lighting

conditions the axis can horizontal in order to detect for example that the image contains a section with reflections. It should be noted that a straight axis could appear curved in the picture due to image

distortion, for example when a camera with a fish-eye lens is applied. In this case the brightness is

preferable evaluated along a curve in the distorted image .

The determination of the spatial brightness

distribution can comprise localization of image areas of a predetermined minimum size, where the brightness is below and/or above a predetermined threshold.

Preferably the correction function is varied, when the image contains at least one boundary section, where the brightness exceeds a predetermined threshold. This boundary section can be located at the upper vertical boundary of the image and contain a bright sky.

Preferably the method is designed to determine, if the image contains at least one bright image section, where the brightness exceeds a predetermined threshold, and at least one dark image section, where the brightness is below a predetermined threshold. The determination of the spatial brightness

distribution comprises the step of dividing the image into lines and/or fields in order to simplify image processing. The fields and/or lines can then be

analysed in terms of the difference between their brightness values according to the gradient method.

The correction function preferably improves the

contrast in tn£

-Linage area, wucreas tiie coriti^'a≤L in the bright area is reduced. This is because in a BSDF situation the dark image areas usually contain the more useful information for the an observer of the image. Preferable the image correction function is a gamma correction function, whereas the slope of the gamma curve is varied depending on the spatial

brightness distribution such that the contrast is improved in dark image areas .

In addition to the contrast correction function a colour correction function can be applied to the image, whereas the colour correction function is varied depending on the spatial brightness and/or colour distribution .

The proposed method can preferably be applied in surroundings -recognition systems for driver assistance systems with cameras which are mounted on vehicles in moving or stationary traffic, in which case the

different areas are defined for example by an upper, relatively bright area which is directed upwards and a lower, relatively dark area which is more important for the evaluation, in the vicinity of the carriageway with relatively dark or shaded objects. The invention can also be applied to static cameras or a scene that is or isn't changing.

The method according to the invention can be performed on a camera system comprising a digital camera, image evaluation means for evaluating the brightness

distribution of the image and correction means for applying a contrast correction function to the image, whereas the image evaluation means are determined for evaluating the spatial brightness distribution of the image and that the correction means are determined for varying the contrast correction function depending on the spatial brightness distribution of the image.

The camera system preferably comprises a digital camera is video camera including a video data generating element that is determined to perform an automatic exposure and/or gain control at the pixel capture level, whereas the video camera is connected to a

Digital Signal Processor (DSP) via a data line and the Digital Signal Processor is determined for evaluating the brightness distribution of the image and for applying the contrast correction function to the image. This is advantageous since a control procedure in the image-evaluation means can avoid automatic control in the camera or be decoupled from it. An automatic control in the camera could otherwise lead to

relatively large changes to the image data and as a result can become unstable. The correction method can preferably operate according to a gamma correction function which respectively selectively amplifies or attenuates the image data of the areas. In this context, the image data can be amplified, in particular in relatively dark areas, in such a way that they produce a higher contrast than the contrast which would result from average brightness of the video image .

It is advantageous that the above mentioned gamma correction function is applied to image data from a video data generating element ( i. g. an imager) whose data were gathered using general automatic exposure and/or gain control mechanisms applied at the pixel capture level .

By achieving all of the above mentioned methods, the camera delivers a picture which, statistically, will have a reasonably good amount information for a general scene. There is normally a gamma in the imager which acts in a sensible manner for general scenes, so the ccrrϊbination of ths imager's default gamma, and AEC/AGC give a good video stream to the digital signal

processing elements.

The important advantage of the invention is that the imager exposure itself is not used to change the DSP gamma. The exposure of the imager simply makes sure that a sensible value is provided to the camera. The gamma controls the relative contrast of adjacent brightness as they are represented in an output data stream and it is this that is changed using the BSDF structure with a bright sky and a darkening foreground (BSDF = Bright Sky Darkening Foreground) .

A principle advantage of this BSDF algorithm is that it doesn't really care what the exposure of the camera is doing but is generally applicable. It is possible to experiment with manipulating the otherwise autonomous behaviour of the cameras for the purposes of some applications, but this is potentially a highly unstable approach. In summary, leaving AEC/AGC to the sensor delivers a body of image data which has structure that may be difficult or impossible for a human viewer to perceive without post-processing, to the DSP and BSDF then makes this hidden structure perceivable to the user by dynamic post-processing of the output gamma of the DSP. The camera system is particularly suitable for a motor vehicle especially as part of a camera based driving assistance systems that takes a video image of the vehicle surroundings and displays the image on screen for inspection by a driver.

The camera system can be connectable to a network, in particular a bus system of the vehicle, and comprise means to receive additional data from the network, if information on if a vehicle light is turned on. This j. IiJ- GiTu at iOn call ϋc UScQ to Vcliiy tiic ilynt iHy

conditions, especially when possible reflections have been detected in the image.

The method according to the invention can be carried out by a computer program or product is proposed which comprises computer-readable program means which are stored on a computer-compatible medium and which, when the computer program product is run on a microprocessor with associated storage means or on a computer, cause said microprocessor or computer to carry out a method or an arrangement for carrying out a previously

described method in a surroundings-recognition system.

Brief description of the drawing

Exemplary embodiments of the invention are presented in the figures of the drawing and explained below. In the drawing : Figure 1 is a schematic illustration of a vehicle on a carriageway with a camera as a component of a

surroundings-recognition system, Figure 2 shows video images from the camera before and after correction of brightness ranges,

Figures 3 to 6 show schematic images with image data analysed according to the invention and

Figure 7 shows a profile of a gamma correction function for correcting the contrast of the image.

Embodiments of the invention

Figure 1 is a schematic image of a diagram in which a camera 1 on a vehicle 2 captures images in a field of view 3. Tile image Cάpturβci oy uric camera ± is image processed by Digital Image Processor and transmitted to a display insides the vehicle's 2 cabin for inspection by the driver.

The image to be captured or the scene has a structure which is referred to below with a BSDF structure, i.e. it has an upper bright, relatively homogenous area 3a and a lower dark, and also relatively homogenous area 3b. In an image-evaluation means 4 in the vehicle 2, the method which is described subsequently can be carried out with different numerical criteria for the areas 3a and 3b.

Figure 2 shows such an image 5a and 5b, each with different numerical criteria. The left-hand image 5a shows the original image data, and the right-hand image 5b shows the image data corrected according to the invention. The image 5a is analysed here, for example, in a vertical raster, which has different brightness values here in the direction from the top to the bottom.

There are now different methods for analysing the image data which will be explained in more detail below. The resulting reliability of three methods is given in the scale 6 with the upper three numbers (100 in each case) . The fourth number 100 below is the resulting overall reliability. The lower number 50 represents the correction intensity, with a quadrant correction, which is known per se, being assumed here.

In the right-hand image 5b in Figure 2 there are also white dots 7 which are intended to represent individual pixels which are to be evaluated, and finally also a brightness histogram 8 is shown top left in the image 5a.

The dots 7 and the histogram 8 are shown here only for the purpose of explanation and are not shown in the application, but rather only used in the image- evaluation process.

Figure 3 shows a representation of the structure of image 5a according to Figure 2 as an image 5c with coarsely screened areas which respectively represent different brightness values. According to a first method, shown in Figure 4, a number of m horizontal fields relating to the upper or lower transgression of a brightness threshold are sampled in a predetermined number of n vertical lines. From both figures it is apparent that in particular the area covering the lower approximately 2/3 of the video image is relatively dark and a correction is appropriate here.

It is also apparent that the method according to Figure 4 operates significantly more precisely with lines than a method with fields since here four adjustable parameters plus the amplification can be set, instead of a similar number plus two 2Ox element patterns.

In the detection methods with fields, the image is therefore divided into fields in a predefined raster and an average brightness per field is determined, so that each field can then be identified and marked as "bright" or "dark" . It is therefore possible to analyse separate patterns for bright and dark pixels with respectively positive results for fields which are to be corrected and negative results for fields which are not to be adapted.

However, the overall pattern of weighted results react more to dark fields than to bright ones, with the result that this method has to be weighted towards dark fields in order to accept scenes in which deep side shadows occur, wnich would otherwise benefit Ixυui correction.

Figure 5 shows an evaluation of fields which are formed, for example, according to the gradient method. This method forms perpendicular gaps between fields and is capable of largely eliminating small changes which are caused by noise or the like, and it generates a cumulative difference. A counter can be increased here for gaps which become cumulatively darker, and can be reduced for gaps which become brighter. If the counter reading exceeds a threshold, this method accepts the image for further processing.

This last-described method can be applied

satisfactorily in images with a dark detailed

background which are often not accepted with other methods. However, images with dark upper areas, such as can occur, for example, in tunnel scenarios, may be rejected. This can often be desirable, but if it is not desired it may often be sufficient to ignore the uppermost areas of the image. However, otherwise the cases in which a bright area is in the centre can be accepted. Figure 5 shows dark pixels as a darkening gradient and bright pixels as a gradient which becomes brighter, and the series at the bottom of the image according to Figure 5 represents the final cumulative gradient. Figure 6 also shows a method with detailed sensing of grey values in the fields of the image in order to permit a positive decision in respect of a correction. This method operates along horizontal series of fields and detects the changes in predefined pixels between adjoining fields, and in this context the number of pixels per field should be at least 40. The differences between the grey values can be classified as follows: small difference = noise = can be ignored moderate difference = can be increased = highly positive moderate difference = can be ignored large difference = corresponds to good contrast = negative extremely large difference = corresponds to extreme bright/dark = positive

Positive results are therefore assigned for correction only if the brightness of at least one of the pixels is below a predefined dark limit. The results which are produced can then be weighted in five horizontal bands, and if the uppermost band is the first, the greatest weighting can be in the third and fourth bands. The described method can function very well if the image which is to be evaluated is a-priori capable of improvement but additional manipulation with a more geometric method is necessary in order to concentrate on a BSDF structure. It is therefore necessary to check continuously whether the potential of the image data is sufficient for a correction. For example, a background with trees could certainly be given a higher weighting but blunt flat objects without a distinctive structure could be ignored. Normally, in these cases an

improvement is aesthetically useful but provides very little information for the driver of a vehicle so that the last-mentioned objects can easily be ignored. In Figure 6, black pixels can be correspondingly weighted as an extreme difference (+1) , and moderately dark grey pixels (+2) and white high-resolution pixels (-1) can be correspondingly weighted. The previously mentioned method is applied to the Y information of the image data and therefore to the brightness control, but it is also alternatively possible to apply it to the colour information of the image data (Cb and/or Cr chroma) , since there are several possible

representations of chroma in video data.

In the second method complex, the correction with which the image data at the output of the camera 1 are further processed, for example by means of a gamma correction function in the image-evaluation means 4, is carried out in such a way that the criteria of the previously described BSDF structure are detected and then a gamma correction is applied to the effect that the image data are weighted by means of the gamma correction function in such a way that the information of the image data in the dark areas of the image can better evaluated and presented. Figure 7 shows such a gamma correction function 10 which is known per se and which represents the change of the Y input signal in relation to the brightness represented in the video signal with respect to a Y output signal. The arrow 11 is intended to indicate how the bend in the gamma correction function 10 is pushed upward as a function of the order of magnitude of the BSDF structure in order to improve the information content of the darker areas .

With the gamma correction function illustrated here, the brightness values corresponding to the Y input signal can therefore be converted in a non- linear fashion. The gamma correction function 10 which is used here differs from the gamma correction function which is usually applied in phosphorescent screens, in particular in cathode ray tubes in television sets, in having greater flexibility. If, for example, adjacent pixels of the video image have brightness values which are very close to one another, those pixels do not provide much contrast; on the other hand, if these brightness values have a large difference, they provide a relatively large contrast. So that the video image is improved in the darker areas, the gradient of the gamma correction function 10 must be increased for relatively low Y input signal values . However, since the overall range of the Y input signal values and the overall range of the Y output signal values have a similar order of magnitude, increasing the gradient of the gamma curve in an area (the

increase in the contrast in an area of the Y output signal corresponds, for example, to a dark area) means that the gradient of the gamma correction function 10 has to be reduced in another range (the contrast in the rest of the video image is, for example, reduced as a result) .

The gamma of the gamma correction function 10 has therefore been changed in order to acquire a larger gradient in the relatively dark areas and a smaller gradient in the middle of the bright areas . For the viewer, a contrast which is closer to the required visual definition and also provides contrast and structure in the relatively dark area means a greater positive effect than the associated negative effect of the reduced contrast in the relatively bright areas .

A significant point here is that the existing knowledge of the scene can be used as part of the correction method to determine the way in which the input video image can be changed to form an improved output video image, Ά further significant poinc is thai, Lhe

of data processing for obtaining the success according to the invention are relatively small. It is possible with the invention for the advantages of such a camera to be used for the image-evaluation process even without employing a high-resolution camera. In the text which follows, a series of further useful applications of the invention will be described by way of example:

In surroundings-recognition systems with cameras, parts of the vehicle are often also captured so that this results in the performance of the surroundings- recognition system being limited. a) When, for example, a puddle on a carriageway is viewed with a downward-directed camera, a headlamp light which is directed downward to a great degree, for example an LED light, can be reflected against the vehicle into a dark scene. This results in the previously described BSDF structure which has the effect that the relatively

insignificant area of the scene (namely the vehicle) is very bright, while the area which is important for the surroundings-recognition system is captured in the camera as a relatively dark area with a relatively small contrast. b) Parts of the vehicle can also be illuminated at night in an application with a camera directed rearward and therefore attenuate details in dark areas . c) Even in an application with a camera which is

facing ahead, the colour and/or other light effects of the vehicle can influence the image data in such a way that they either become too bright Oi^" too dark. If it is too dark, there are possible ways of improving the bright scene information such as a sunlit carriageway behind the visible edge of a shadow thrown by the

vehicle . d) In an application with a camera which is facing sideways, there may also be problems with the imaging if, for example, a front scattered light excessively brightens an area of the scene so that the relatively dark areas of the scene which are, however, important for the image-evaluation process can be perceived less. e) Generally, in cameras which are facing ahead there are problems, for example when entering a tunnel or when entering a darkened garage entrance, in which case the dark area is surrounded by the bright area and nevertheless at the same time an attempt is made to obtain the optimum information content of the image data. In such straight-ahead ~ views, the image information in the area which is being entered is usually the most useful. f) If a vehicle is travelling on a wet surface with strong sunlight which is incident at a low angle, the effect of such strong light can be reduced with a suitable heuristic. In conjunction with a surface film, the light can also be polarized in order to improve the effect. g) In all cases when there is an area of the vehicle which can be seen by the camera and which has such a light- influencing effect, it is possible to achieve success with the method according to the invention. In general, it is therefore possible to apply the invention advantageously wherever there is some part of the vehicle which is within the field cf vision of the camera but outside the cuied which is of interest for the image-evaluation process.

For example, if a camera which is directed

rearward sees, at night, the illumination of a car registration plate in its field of vision, with the method according to the invention the effect of this illumination can be reduced in the image- evaluation process .

The method according to the invention may also be helpful if the side of a bright vehicle is bright in a camera which is facing sideways or if a dark vehicle is in relatively dark surroundings in daylight . - The colour of the vehicle can also already be

input into the video-based surroundings- recognition system during the fabrication process so that this can automatically be taken into account in the image-evaluation process in the driving mode. This is particularly advantageous when it is known during the image-evaluation process where the vehicle can be found in the field of vision of the camera. h) There is a natural tendency with the wide-angle lens systems for so-called optical fall-off to occur in the direction of the lower edge of the image. Some video systems take this into account by artificially increasing the brightness of the scene when moving in the radial direction away from the centre of the image. In this context, a heuristic corresponding to a BSDF effect can be applied using the method according to the

invention. For example, it may possibly not be desired to increase the brightness of the areas of of the camera but are not at an extreme angle in an area which is of interest for the image- evaluation process.

i) The method can also be used as a countermeasure to the oversaturation due to a strong light source in a relatively dark scene such as that from a vehicle lamp at dusk.

The concept according to the invention can also be applied to scene information which requires so-called white balance, for example in order to calculate the degree of white balance through knowledge of the colour of the vehicle so that the degree of white balance can be reduced. In general, a video system calculates the image data in such a way that the scene contains a similar proportion of the colours red, green and blue in order to reduce the effect of the colour of the lighting element on the illuminated objects, and therefore to ensure that the colours which are seen by the user represent the object more than the light source. This is a so-called "grey world assumption" and functions in an extremely acceptable way for most scenes. However, if the object in the video system can be seen in the way in which it has been shown in the previously described applications, and the object is a significant proportion of the colour red in the field of vision, the red proportion must be suppressed in the case of a relatively neutrally illuminated element in the field of vision or the blue and green proportions are enhanced in order to preserve the "grey world assumption" . If the level of the proportions is

therefore known a-priori, the incorrect control of the white balance can be reduced. In the abovementioned examples, the respective position of the property which is to be changed is known. In particular in the case of the examples a) - c) it is which side of the vehicle is being viewed by the camera, for example either by virtue of the

manufacturer's documents or due to a presetting during the manufacture of the vehicle. It is therefore

possible to carry out specific treatment of the image data using the configurable parameters.

In principle, the gamma correction functions which have been described above can be amplified or applied more specifically by means of the known vehicle colours. For example, in case d) it is possible to make a good assumption about the range of the incident light.

Furthermore, information indicating whether the light is switched on can also be obtained from the vehicle data in order to improve the reliability of the

measures which are to be performed.

The invention can be applied not only to vehicles in road traffic but also to other image-evaluation

processes to which different brightness ranges in a scene which is to be evaluated generate unsatisfactory image data. With the proposed gamma correction function it may also be possible to bring about compensation of suboptimum gamma values which are frequently found to occur in image systems/displays in vehicles because image systems often operate with obsolete gamma values of the manufacturers of the displays .

Claims

1. Method for improving the contrast of a camera image comprising an evaluation of the brightness distribution of the image and application of a contrast correction function to the image,

characterized in that

evaluation of the brightness distribution includes determination of a spatial brightness distribution of the image, whereas the contrast correction function is varied depending on the spatial brightness

distribution.

2. Method according to Claim 1, characterized in that determination of the spatial brightness distribution includes determination of the brightness at

predetermined image locations .

3. Method according to Claim 2, characterized in that determination of the spatial brightness distribution includes determination of the brightness distribution of the image along at least one predetermined axis and/or geometric curve.

4. Method according to Claim 2 or 3 , characterized in that determination of the spatial brightness

distribution comprises localization of image areas (3a,- 3b) , where the brightness is below and/or above a predetermined threshold.

5. Method according to Claim 4, characterized in that the correction function is varied, when the image contains at least one boundary section (3a) , where the brightness exceeds a predetermined threshold.

6. Method according to Claim 5, characterized in that the boundary section (3a) is located at the upper vertical boundary of the image.

7. Method according to Claim 4, 5 or 6 , characterized in that the image contains at least one bright image area (3a) , where the brightness exceeds a predetermined threshold, and at least one dark image area (3b) , where the brightness is below a predetermined threshold.

8. Method according to one of the preceding claims, characterized in that determination of the spatial brightness distribution comprises dividing the image into lines and/or fields.

9. Method according to Claim 8, characterized in that the fields and/or lines are analysed in terms of the difference between their brightness values according to the gradient method.

10. Method according to Claim 8, characterized in that in order co recoxu yxey values in the fields of the video image in a detailed fashion, the changes at predefined pixels are analysed between adjacent fields, wherein the differences of the grey values are

evaluated in such a way that a correction is performed only if the brightness of at least one of the pixels is below a predefined dark limit.

11. Method according to one of the Claims 4 - 7, characterized in that the different areas (3a, 3b) are defined by an upper relatively bright area (3a) which is directed upwards and a lower relatively dark area (3b) which is more important for the evaluation, in the vicinity of the carriageway with relatively dark or shaded objects.

12. Method according to one of the Claims 4 - 7, characterized in that the different areas are defined by a relatively dark, inner area, in particular by tunnel entrances or garage entrances, which is more important for the evaluation, and by a surrounding, relatively bright area.

13. Method according to one of the Claims 4 - 7, characterized in that the different areas are defined by a relatively bright area due to reflections or salient colouring, and a relatively dark area which is more important for the evaluation.

14. Method according to Claim 13, characterized in that the areas which are relatively bright due to reflections or salient colouring are parts of the vehicle (2) carrying the cameras (1) .

15. Method according to one of the preceding claims, characterized in that the correction function does improve the contrast in the dark image (3b) areas, ^ωherea.s the contrast in the bright αrccic; (3Ξ.) is reduced.

16. Method according to one of the preceding claims, characterized in that the correction function is a gamma correction function, whereas the slope of the gamma curve is varied depending on the spatial

brightness distribution.

17. Method according to one of the preceding claims, characterized in that a colour correction function is applied to the image, whereas the colour correction function is varied depending on the spatial brightness and/or colour distribution.

18. Camera system comprising a digital camera (1), image evaluation means for evaluating the brightness distribution of the image and correction means for applying a contrast correction function to the image, characterized in that the image evaluation means are determined for evaluating the spatial brightness distribution of the image and that the correction means are determined for varying the contrast correction function depending on the spatial brightness

distribution of the image.

19. Camera system according to claim 18, characterized in that the digital camera (1) is video camera

comprising a video data generating element that is determined to perform an automatic exposure and/or gain control at the pixel capture level, whereas the video camera is connected to a Digital Signal Processor (DSP) via a data line and the Digital Signal Processor is determined for evaluating the brightness distribution of the image and for applying the contrast correction function to the image.

20. Camera system according to claim 19, characterized in chac uhe camera system ia ut;Lfc;xiiLixit;d LOJL a motor vehicle (2) .

21. Camera system according to claim 20, characterized in that the camera system is connectable to a network of the vehicle, whereas the camera systems comprises means to receive data from the network about the status of at least one vehicle light.

22. Computer program product which is stored on a computer-usable medium and comprises computer-readable program means which, when the computer program product is run on a microprocessor with associated storage means or on a computer, cause said microprocessor or computer to carry out a method according to one of the claims 1 to 17.