METHOD FOR GENERATING VIDEO DATA STREAM

Abstract

The invention relates to a method for generating video stream data relating to different fields of vision from a camera having a sensor which records high resolution images in a distorted manner. According to the invention, image corrections are performed for separate fields of vision, and a total video stream is output from the differently corrected image data.

Description

The present invention is concerned with what is claimed in the preamble, and thus relates to the use of cameras.

Video cameras are known. They comprise a sensor that repeatedly records images in order to be able to provide image data for a display or for storage. High-resolution cameras, for example having fish-eye lenses, are also known in which the lens-governed optical distortion can be compensated for by data processing for display purposes.

Video streams make high requirements of transmission lines, data-processing processors and display devices. On account of bandwidths or computing capacities, it is often not possible to enable images to be represented with high resolution and/or jointly. As cameras become used more widely, however, users increasingly want to be able to obtain video streams from a specific camera even when a powerful computer system is not currently available. By way of example, mention shall be made here of the desire of many subscribers to be able to use video telephones with low outlay or to ensure monitoring via a mobile telephone using web-based cameras; this allows parents, for example, by means of cameras fitted accordingly at home, to see their children even when they are absent, and to check whether their children are fast asleep at night. Such checking may be expedient within the home or near home, in a manner comparable with the “babyphones”, which hitherto have only had acoustic action. The fact that other surveillance tasks may likewise be expedient in addition, for instance for the surveillance for industrial installations, outdoor installations, etc., shall be mentioned only for the sake of completeness.

What is problematic, however, is that the display of an individual video channel is often insufficient, for instance because a relatively large region or a number of children is/are to be monitored. Although it would be possible, in principle, to progressively interrogate a plurality of cameras from which individual images are obtained cyclically, this is inadequate precisely in the case of danger and does not meet the requirements of many users. Moreover, the network load is high.

It is desirable, therefore, to enable the reproduction of video streams which can show many different individual images even on less powerful devices.

The problem addressed by the present invention consists in providing novel subject matter for industrial application.

The solution to this problem is claimed in independent form. Preferred embodiments are found in the dependent claims.

Consequently, the present invention proposes, in a first basic concept, a method for generating video stream data relating to different fields of view from a camera having a sensor which records high-resolution images in a distorted manner, it being provided that a respective image correction is performed for separate fields of view, and a total video stream is output from the differently corrected image data.

Consequently, one essential aspect of the invention is based on the insight that it is possible not only to determine separate excerpts from a large quantity of image data and, if appropriate, to perform for said excerpts image corrections which enable adaptation to a simple display device, rather that it thus also becomes possible to provide a single total video stream carrying different fields of view combined in itself in order thus to enable display on a display device without a high computational load. In this case, it is possible without problems to provide the total video stream as a conventional video stream such as, for example, with MPEG coding, etc., to enable compression, etc.

Moreover, the total video stream can be generated from original images stored at or in the camera. The storage only of the original images at the camera requires less storage space in comparison with a case in which rectified excerpts also additionally have to be stored. Image data can be stored at the camera by being saved to a nearby hard disk via a USB interface, for example.

It suffices, in particular, to safeguard an area using just one camera according to the invention; thus, by way of example, a plurality of checkouts in a supermarket can be monitored by means of a single camera.

It is preferred that the camera is a wide-angle camera, in particular a fish-eye camera, which covers a half-space or almost a half-space, that is to say has a viewing angle of 170° or higher. In this case, particularly great image deformation, that is to say distortion of the high-resolution images, is to be expected particularly at the image edge and, moreover, the separate image connection for different fields of view is particularly expedient and necessary. Thus, an image rectification which is necessary for a field of view lying at the outer left image edge will have to appear significantly different than, for instance, for a field of view lying near the central viewing axis. It becomes possible, by means of separate rectifications for example using algorithms for which only different parameters have to be predefined, to obtain a geometry that gives a natural impression for different image regions, which makes continuous observation more pleasant.

The high-resolution images are typically recorded as individual images by the sensor and combined to form a video stream only in the camera. The images can originate from a megapixel sensor, having a resolution of two to twelve megapixels, for example. Since, in this case, it is not necessary, at any rate not absolutely necessary, to convert the entire image information from all pixels of the high-resolution image to form a video stream, a typically sufficient computing power for performing the method without particularly measures is available even with conventional cameras.

In one particularly preferred variant, the correction of the images and the combination of the corrected images to form a total video stream will therefore be effected within the camera. This is also advantageous because it can thus be ensured that the individual partial images of the total video stream are based on the best possible sensor information, without losses of information being brought about by compression for data transmission prior to the geometrical rectification. Detail reproduction with good resolution is thus ensured despite the use of a total video stream with a low data rate.

The preferred camera of the invention accordingly comprises a sensor, a processing unit for correction and combination and a communication unit. No separate additional devices are required, although such additional devices could indeed plausibly be provided, for instance if older cameras having low computing power with very high-quality optics should be used or an already installed camera that is still weak computationally is not intended to be immediately replaced on account of particularly costly vandalism protection against destruction.

The method of the present invention can be performed such that virtual cameras are defined within the total field of view. Said virtual cameras can, in particular, be panned, tilted or zoomed like real cameras, such that an intuitive control of the fields of view is possible, without wearing parts having to be moved. Moreover, it is not evident to observed persons whether they are currently the focus of attention. This increases the deterrence. The corrections can then be performed for the virtual, separate fields of view obtained with the virtual cameras. In this way, only those image pixels which lie in the field of view of a virtual camera have to be subjected to treatment. Typically, the fields of view of differential virtual cameras will be situated separately, that is to say will not adjoin one another. It should be mentioned, however, that overlaps can also occur in exceptional cases.

The virtual cameras can preferably be individually panned, tilted and/or zoomed. This can preferably be done by the manipulation of a pointing device on the partial image which corresponds to the virtual camera in the total video stream. A typical pointing device manipulation is, for example, a mouse actuation or a joystick actuation.

In one particularly preferred variant, on the one hand a panoramic image and on the other hand at least one image excerpt can be provided. The panoramic image, particularly when a fish-eye camera is used, can be cropped at the upper and lower regions and need not extend as far as the extreme outer left and right image boundaries. This ensures a sufficiently high resolution in all regions. It should be mentioned, however, that this is not of essential importance in the case of display devices having a low resolution per se, such as mobile telephones.

In typical applications, non-overlapping fields of view will be defined.

The image correction can be carried out, firstly, in order to bring about a geometrical rectification such as is particularly necessary in the case of fish-eye lenses. At the same time, however, it is also possible to correct brightnesses, color hues, etc. For this purpose, it can be taken into consideration that specific regions of an image either have a different brightness from the outset on account of the conditions in the observed fields of view and/or a specific darkening occurs in the edge regions. The same applies to color corrections which may be necessary firstly as a result of differently tinted illuminations of different field of view regions and secondly as a result of lens aberrations. In this case, the image correction is typically performed in such a way that when the individual fields of view are combined to form a total video stream, this results in a video stream which at least approximately has the same brightness for all the fields of view, and which will preferably also have the same tint. This makes it easier to view the images on a display device—such as a mobile telephone—that is inexpensive and thus typically has only a low dynamic range and color gamut.

The total video stream is typically output to a display device exhibiting single-channel reception. In this case, single-channel does not relate to whether signals are actually received only on a single channel—it is pointed out that in modern mobile telephones, for instance, a plurality of frequencies are used simultaneously for data transmission—rather reference is made to the fact that the display device will typically be designed only for receiving a single video stream. Such a video stream will typically not constitute particularly high loading for the display device. Since the rectification will not be effected at the display device either, this results overall in low loading.

The video stream can be transmitted in a standard video stream format such as, for example, 16:9 or 4:3 or the like. The fact that here different compression methods such as MPEG, such as are possible and known per se in the prior art, can be employed should not only be mentioned but, with regard to the fact that the application will be particularly efficient, even emphasized.

It will typically be possible and preferred if a plurality of display devices can be addressed from one camera, in particular by the provision of video streams which are in each case conditioned in accordance with the present invention; the fact that instead individual or all display devices can obtain the full data video stream should be mentioned. Typically, a camera will be able to rectify m field of view and be able to output video streams to n consumers, where it will be the case that m≧2 and n≧1.

If different total video streams are provided for different display devices, it can happen that at least one field of view is the same in the different total video streams or one of the fields of view in one of the video streams is completely encompassed in a field of view of the other total video stream. In such a case, it is respectively preferred if, for this field of view, the image correction has to be performed only once. In this case, the image correction which is performed only once can be, if appropriate, a geometrical image correction, particularly when the fields of view largely overlap; other necessary corrections such as hue or brightness adaptation can, by contrast, refer to other image constituents of the total video stream respectively to be provided for a display device. Even in such a case, as a result of the common reference to a single (geometrical) image correction, this then results in a reduction of computing load at the camera.

A particularly preferred display is obtained if at least one panoramic image is combined with at least two detail images to form a total video stream and/or four individual images are combined to form a total video stream. It is advantageous for all the partial images of the total video stream to have the same timing, that is to say to be synchronous. This is particularly advantageous if sound reproduction is intended to be effected simultaneously.

The invention allows a plurality of cameras, which have had to be used heretofore for specific tasks such as gasoline pump monitoring at a relatively large gas station or checkout monitoring of a plurality of checkouts in a supermarket, to be replaced by a single high-resolution camera. This reduces the installation outlay and the maintenance costs, the network capacity utilization remaining low according to the invention. It is advantageous in particular in this case if the camera can identify objects and enables automatic tracking of said objects in the partial images of the total video stream, for which purpose and in which case in particular the settings of virtual cameras can be changed, such as rectification, image brightness, tilt, pan direction and/or zoom factor.

The invention is explained below merely by way of example with reference to the figures. In the figures:

FIG. 1 illustrates an image representing a scene recorded with high resolution by a fish-eye camera,

FIG. 2 a illustrates an image region rectified for representing a panorama, said image region representing a first excerpt from FIG. 1,

FIG. 2 b illustrates the image region from FIG. 2a with different rectification,

FIG. 3 illustrates a different excerpt from the scene recorded by the fish-eye camera,

FIG. 4 illustrates a further excerpt from the scene,

FIG. 5 illustrates another excerpt from the scene,

FIG. 6 illustrates a further corresponding excerpt from the scene, and

FIG. 7 illustrates an image of a composite video stream comprising the panoramic overview from FIG. 2, the excerpt from FIG. 5 and the excerpt from FIG. 6.

The present invention relates to a method for reproducing video stream data relating to different fields of view from a camera having a sensor which records high-resolution images in a distorted manner. One typical image of this type is illustrated in FIG. 1. It should moreover be pointed out that where the figure designation such as “FIG. 1” or “FIG. 5” has been copied into the image, this is clearly not part of the originally recorded image, but was necessary for explanation reasons.

It is assumed in the present case that the general technology of megapixel cameras is known to a person of average skill in the art, and likewise the algorithms for carrying out image rectification in order, for instance, to make the counter in FIG. 1 appear less curved, cf. FIG. 2 partially rectified in this respect. Therefore, they will not be discussed per se below. In the present case, the camera is preferably designed, however, such that a given image excerpt, for example that from FIG. 2a, can be rectified in different ways. This is deemed to be novel and advantageous. This is illustrated in FIG. 2b, which in the case of continuous monitoring of an area by means of a panoramic image conveys a more natural impression of the area. However, the possibility of using a plurality of rectifications is not mandatory, but rather only greatly preferred.

It is furthermore assumed that the technique for the definition of excerpts and the rectification of individual excerpts such as the excerpts from FIG. 3, FIG. 4, FIG. 5, FIG. 6 are also known to the person of average skill in the art insofar as a specific field of view has to be defined within an image. Said field of view will then be rectified in a dedicated way in a manner known per se, in which case different rectifications can once again preferably be available.

In this case, it should be pointed out by way of example that the shopper at the outer right image edge is represented with greater distortion in FIG. 3 than in FIG. 5, which shows the same excerpt with different rectification.

Rectification can plausibly be performed in a different way, for instance in such a way that edge regions are rectified to a lesser extent or instead to a better extent. It is preferred, therefore, to offer different rectifications and to allow the user to choose a rectification which rectifies particularly well regions that are important to said user and/or conveys to said user an impression that is particularly pleasant to said user because it seems natural; it is thereby possible, for example, for specific lines which would otherwise be curved incorrectly in the image to be represented straight.

In order, then, proceeding from an image in accordance with FIG. 1, to arrive at a video stream according to the invention corresponding to FIG. 7 with one panoramic section and two detail sections which are combined to form a joint video stream in a standard format, the respective individual excerpts being renewed simultaneously with one another, such that a high degree of synchronism in observation can be obtained, the following procedure is adopted:

Firstly, a first excerpt, for example the panorama from FIG. 2, is defined. This can take place at the display device in a video stream transmitted as a total video stream in accordance with the images from FIG. 1 in a largely unrectified, but compressed image. The display means therefore do not have to rectify the image at any point in time.

The excerpts can be defined in accordance with a pan-tilt-zoom movement in the case of a conventional pan-tilt-zoom camera, with the stipulation that an image aspect ratio can additionally be predefined as well.

After the definition of a first image region that a user wants to see on an inexpensive display device, it is possible in addition to define a further excerpt, for example as illustrated in FIG. 7, and then a third excerpt as illustrated in FIG. 5, for example. It is possible to provide an automatic arrangement and predefinition of excerpts. In such a case, the user does not have to be concerned with the image aspect ratio, but rather only has to correspondingly place respective excerpts into the image from FIG. 1.

After the individual excerpts have been defined, excerpts are respectively taken from the non-rectified image data in accordance with FIG. 1 and are rectified per se and then a brightness adaptation of the here three image excerpts to one another is performed in order to achieve a uniform illumination of all the display regions of a display device that receives and represents the video stream. The fact that, if appropriate, with provision of in each case a plurality of different rectification possibilities for the different image regions, the rectifications of the individual images are adapted to one another in order to ensure the best possible, relaxed viewing, should be mentioned as preferred. After geometrical image rectification, contrast adaptation, color balancing, etc., the individual images are then combined with one another to form an image in a standard format such as 16:9 in accordance with FIG. 7. The excerpts corresponding to progressively successive images in accordance with FIG. 7 then constitute a video stream, in each “window” a video film running which, moreover, is highly synchronous with the other images, which is advantageous for sound reproduction. The video stream can be compressed and output via a single channel to a display device, where it can be represented with low computational complexity. The network load as a result of the transmission of the image is low in this case.

Claims

1. A method for generating video stream data relating to different fields of view from a camera having a sensor which records high-resolution images in a distorted manner, characterized in that a respective image correction is performed for separate fields of view, and a total video stream is output from the differently corrected image data.

2. The method as claimed in the preceding claim, characterized in that the camera is a wide-angle camera, in particular a fish-eye camera having a viewing angle of greater than 170°.

3. The method as claimed in either of the preceding claims, characterized in that individual high-resolution images are recorded by the sensor in a distorted manner.

4. The method as claimed in any of the preceding claims, characterized in that the respective image correction and combination to form a total video stream are effected within the camera.

5. The method as claimed in any of the preceding claims, characterized in that virtual cameras are defined for the definition of the separate fields of view, and the total video stream is combined from the virtual individual camera video streams of the virtual cameras.

6. The method as claimed in the preceding claim, characterized in that, for each virtual camera, an independent pan, tilt and/or zoom change can be performed, in particular by a change on the corresponding partial image from the total video stream.

7. The method as claimed in any of the preceding claims, characterized in that at least one panoramic image and one image excerpt are provided with the fields of view.

8. The method as claimed in any of the preceding claims, characterized in that at least two non-overlapping fields of view are defined.

9. The method as claimed in any of the preceding claims, characterized in that an image correction is performed with regard to image brightness, geometrical rectification and/or image resolution.

10. The method as claimed in any of the preceding claims, characterized in that the total video stream is output to a single-channel display device.

11. The method as claimed in any of the preceding claims, characterized in that the video stream is output in a standard image format or standard video stream format.

12. The method as claimed in any of the preceding claims, characterized in that a plurality of display devices are addressed by one camera, in particular in each case in the manner according to the invention.

13. The method as claimed in the preceding claim, characterized in that image data for different display devices are combined to form different total video streams and at least one field of view is the same in the two different total video streams, the image correction for this field of view being performed only once.

14. The method as claimed in any of the preceding claims, characterized in that at least one panoramic image and two detail images and/or four individual images are combined to form a total video stream.

15. The method as claimed in any of the preceding claims, characterized in that the type of geometrical image correction performed in the camera, preferably performed per virtual camera, can be determined by a user.

16. The method as claimed in any of the preceding claims, characterized in that in the camera objects are identified and the identified objects are tracked by means of virtual cameras in order to be able to represent the objects thus tracked in the total video stream in partial images.

17. The method as claimed in any of the preceding claims, characterized in that the total video stream generated is output for display on a display device.