The invention relates to a method and a device for editing recorded images from a digital video camera.
Digital video cameras with electronic image sensors for moving images are utilized in many areas of film and TV production. They contain one or more electronic image sensors and apply different sensor technologies, such as CCD or CMOS, when the electronic image sensors have different sizes.
Since it is easier to produce smaller electronic image sensors, digital video cameras equipped with these image sensors are more widely available. Here, an accepted disadvantage during use of these digital video cameras is that, inter alia, there is an increase in depth of field in the recorded object as a result of the small electronic image sensor. In many productions, this effect is undesirable because a small depth of field gives the option to the cameraman of directing the viewer's attention to a particular plane, e.g. to the face of an actor. The cameraman loses an essential style device, if the depth of yield is too big.
Specific parts of a film production are recorded using both a motion picture cine camera with a cine lens and a digital video camera. However, if the electronic image sensor in the digital video camera is smaller than the camera aperture of the digital video camera, the recorded scenes cannot be cut together because the respective angles of view do not fit together. For this reason, it is desirable for cine lenses used in motion picture cine cameras also to be used in digital video cameras.
However, using cine lenses for motion picture cine cameras in digital video cameras does not remove the aforementioned problem, which originates purely from the size of the electronic image sensor, because an optical adaptation using only imaging optics does not remove the disadvantage relating to a depth of field that is too large.
As per
In this arrangement, the ground-glass screen 2 decouples the two optical systems: the cine lens 1 on the one hand and the relay optics 3 or sensor assembly 4, 5 on the other hand. An analog/digital converter is part of the sensor assembly 4, 5; however, it is not illustrated separately in the schematic illustration as per
However, two new problems arise when solving the problem of also being able to use small electronic image sensors by decoupling the two optical systems by means of a ground-glass screen.
Firstly, the structure of the ground-glass screen used in the beam path of the digital video camera can be recognized in the image produced by the digital video camera, with the recognizable structure of the ground-glass screen becoming ever more visible as the stopping down of the cine lens increases. The use of a finer grain for the ground-glass screen does not yield an improvement because this would lift the decoupled state between the two optical systems in the digital video camera.
Secondly, there is vignetting in the edge region of the images recorded by the digital video camera as a result of non-matched pupil positions between the cine lens and the relay optics. Here, the amount of vignetting is dependent on the utilized cine lens and the position of the exit pupil. Although the keyhole effect caused by the vignetting is suppressed by the ground-glass screen, it is not removed completely, with the strength of the effect of the vignetting being dependent on the respectively utilized type of cine lens and on the lens aperture of an iris diaphragm of the cine lens.
In order to remove the ground-glass screen structure, DE 20 16 183 B has disclosed the practice of making a ground-glass screen oscillate rapidly within its areal plane; the grain structure of the ground-glass screen is smeared as a result of this. However, this method for removing or reducing the grain structure of the ground-glass screen is afflicted by the disadvantage that the design of the oscillation-generation device is very complicated due to the mechanically moved ground-glass screen because of the requirement that the ground-glass screen may only deviate by a few hundredths of a millimeter from its areal plane despite the fast oscillatory motion; otherwise the images recorded by the digital video camera become blurred.
Moreover, the fast oscillatory movement is connected to noises, with, in an implemented embodiment, the ground-glass screen rotating in its areal plane normal to the external surfaces since, in mechanical terms, this leads to the simplest mount. However, this leads to corresponding Coriolis forces when the video camera undergoes a panning motion, which Coriolis forces in turn put a load on the bearings of the rotational apparatus and/or influence the position of the ground-glass screen for the duration of the panning motion.
In any case, the vignetting problem is not removed by the rotating or oscillating ground-glass screen.
The document YU W: “PRACTICAL ANTI-VIGNETTING METHODS FOR DIGITAL CAMERAS” IEEE TRANSACTIONS ON CONSUMER ELECTRONICS, NEW YORK, N.Y., US, Volume 50, Number 4, November 2004 (2004-11), pages 975-983, XP001224730, ISSN 0098-3063 has disclosed a method for automatically correcting vignetting errors in images from a digital video camera; here, a reference image is recorded by the video camera and a correction factor is calculated for each pixel position, which results in a correction image that corresponds to the calculated correction factors in pixel values. Hereafter, images recorded by the same video camera are corrected by multiplying them by correction factors that are stored in a table. Missing correction factors are calculated by interpolation by applying hyperbolic-cosine functions.
This correction method requires very high computational intensity, both in generating the correction images and in correcting the recorded images and is not suitable for using cine lenses because the angles of view do not change.
The object on which the present invention is based is to specify a method and a device for editing recorded images from a digital video camera of the type mentioned at the outset, which, with little hardware and software intensity, allow the use of cine lenses without adversely affecting the image quality, even in conjunction with digital video cameras with a small electronic image sensor.
The solution according to the invention specifies a method and a device for editing recorded images from a digital video camera, in which a cine lens can also be used in conjunction with a digital video camera with a small electronic image sensor without this adversely affecting the image quality, in particular by the grain structure of an imaging disk, more particularly of a ground-glass screen or of a fiber plate, arranged in the beam path of the digital video camera becoming visible or by the occurrence of vignetting effects, wherein low hardware and software intensity is required for achieving a high image quality.
The solution according to the invention assumes decoupling between the optical systems, to be precise between the projection on the imaging disk of a recorded image using the cine lens on the one hand and the projection on the electronic image sensor of the imaging-disk image using relay optics on the other hand, and electronic correction of the image errors caused by using the cine lens and the imaging disk.
To this end, the digital video camera records a calibration image used to calculate correction values both for the grain structure on the imaging disk and for the vignetting in the edge region of the recorded images, and these correction values are linked after the calibration to the recorded images in the recording mode.
According to a further exemplary feature of the invention, since the effects of the grain structure on the ground-glass screen or fiber plate and the vignetting are dependent on the lens aperture of the cine lens, n vignetting and structure matrices are established for n different lens apertures of the cine lens. In the process, the following is taken into account: although the grain structure is independent of the lens aperture of the cine lens, the grain structure becoming visible ultimately is dependent on the lens aperture of the cine lens because if the lens aperture of the cine lens is small the light beams emerging from the cine lens impinge in parallel on the ground-glass screen such that the grain structure of the ground-glass screen is clearly visible, while in the case of large lens apertures the light beams emerging from the cine lens impinge on the ground-glass screen at different angles, and so the grain structure is not as clearly visible. As a result of this—albeit small—dependence of the grain structure on the lens aperture of the cine lens when considering the vignetting effects, the grain structure is also established at different lens apertures of the cine lens.
Since the vignetting effects and grain structures are also dependent on the respectively utilized lens type, the vignetting and structure matrices are, according to a further feature of the invention, respectively established for a particular cine lens type and used as correction values for the images from the digital video camera recorded during the recording operation.
The grain structure of the imaging disk is established when the imaging disk is installed. For this purpose, arranged in the beam path of the digital video camera are the cine lens, the imaging disk embodied as a ground-glass screen or fiber plate, relay optics and the electronic image sensor apparatus for generating one or more correction images used to calculate the correction matrices for the grain structure of the imaging disk and the vignetting effects, the individual values of which then serve in the recording mode of the digital video camera for correcting the recorded images.
The recorded images from the recording mode of the digital video camera can be corrected electronically either on the level of the raw sensor data or in the RGB color space after image processing.
In order to establish the correction matrices, correction values are generated for each pixel in the calibration image, by
This method for establishing and processing the correction matrices by averaging over adjacent regions, for example over a block of 36 or 49 pixels, leads to very good results but requires increased computational intensity. For simplification purposes, it is also possible to obtain only local mean values of the brightness for a prescribable number of pixels, e.g. 20 pixels, in a current line for both the vignetting and the structure matrix.
The correction matrices can either be calculated from the calibration images in a data processing unit of the digital video camera and said correction matrices can be correlated to the recorded images in a recording mode of an image processing unit of the digital video camera, or the calibration images recorded by the digital video camera are output as a video signal at an image output of the digital video camera and transferred to an external PC, in which the correction matrices are calculated and returned to the image processing unit in the digital video camera via a data interface and said correction matrices are correlated to the recorded images in a recording mode of the digital video camera.
In the recording mode of the digital video camera, the recorded images are corrected in a real-time capable system, for example in a programmable logic component (FPGA—field programmable gate array), in which each individual pixel in the recorded image is firstly multiplied by the same pixel in the vignetting matrix and then multiplied by the same pixel in the structure matrix such that those points that are too dark as a result of the ground-glass screen structure or vignetting have the individual pixels multiplied by a factor greater than 1 and hence they are brightened, while those points that are too bright as a result of the grain structure of the ground-glass screen or the vignetting are multiplied by a factor of less than 1 and hence they are darkened, and so this results in a consistent recorded image that is free from the influences of the grain structure and vignetting.
In order to take into account the lens aperture of the respectively utilized cine lens, the values of the lens aperture of the cine lens are detected by a sensor connected to the cine lens and stored together with the correction values of the pixels.
If the cine lens does not have appropriate sensors such that it is not possible to enter electronically the values of the lens aperture of the cine lens, both the correction factors and the switching between the different correction matrices, recorded dependent on the lens aperture of the cine lens, can also be entered manually.
The device according to the invention for editing recorded images from a digital video camera with a cine lens for projecting recorded images onto an imaging disk, arranged in the beam path of the cine lens, of the digital video camera, an electronic image sensor apparatus and an image processing unit with
The idea on which the invention is based is to be explained in more detail on the basis of exemplary embodiments illustrated in the drawing.
In the arrangement of components of a digital video camera described previously, an image sensor that is smaller than the image field of the digital video camera may be used for converting the moving recorded images. However, it is possible to identify the structure of the ground-glass screen in the image of the digital video camera in this arrangement; and this always increases with increasing dimming of the cine lens, i.e. as the size of the lens apertures decreases. Moreover, vignetting can be identified in the edge region of the recorded images as a result of non-matched pupil positions between the cine lens and the relay optics, particularly if the cine lenses are made by different manufacturers and hence the position of the exit pupil is not defined in any way. Although the keyhole effect created by the vignetting is reduced by the ground-glass screen, it is not completely removed.
In order to remove the grain structure, which is caused by the ground-glass screen, and the vignetting effect, the arrangement of a digital video camera illustrated in
Alternatively, it is possible that only the time-critical part is embodied as a component of the digital video camera while a processor for calculating correction matrices is arranged externally.
In a further alternative, the entire image processing unit 7 can additionally be attached outside of the actual video camera, and so there is no need to interfere with the actual video camera in order to use the solution according to the invention.
The image processing unit 7 contains a controller or computer 72, which edits the raw or RGB data 71 output by the sensor assembly 4, 5 and/or outputs the raw or RGB data 71 and is connected, on the input side, to an external data interface 70 and to a buffer memory 73, to which, on the input side, the raw or RGB data 71 is applied. On the output side, the controller/computer 72 is connected to both a vignetting-matrix memory 74 and a structure-matrix memory 75. The output of the vignetting-matrix memory 74 is connected to a first multiplier 781, to which, additionally, a first correction factor 72 is applied and, on the output side, is routed to an input of a second multiplier 782, with the raw or RGB data 71 being routed to the second input of the latter. The output of the structure-matrix memory 75 is routed to a first input of a third multiplier 783, wherein a second correction factor 77 is applied to a second input and the output thereof is routed to a first input of a fourth multiplier 784, the second input of which is connected to the output of the second multiplier 782 and on the output of which a corrected recorded image 79 is output.
At least one calibration or correction image is generated with the aid of the design of a digital video camera illustrated in
Since the effects of the grain structure and the vignetting are dependent on the lens aperture of the cine lens 1, a plurality of correction matrices are established at different lens apertures of the cine lens 1 for the grain structure of the ground-glass screen 2 and for the vignetting. Moreover, in the case of different types of cine lenses, a plurality of correction matrices are produced for the different lens apertures together with a specification of the lens type.
The following text describes the method according to the invention for editing recorded images from a digital video camera or the function of the image processing unit 7 illustrated in
The flowchart illustrated in
If the local brightness of the correction image is brighter than the mean of the overall image, the ratio
Imean overall image/Ilocal current mean
results in a value of less than 1. If the current value of the local brightness of the correction image is darker than the mean of the overall image, the aforementioned ratio results in a value of greater than 1. The vignetting matrix formed in step e and illustrated in a graph in
The deviation of each pixel from the local average is established in step f. This generates an image of the ground-glass screen structure in the form of a structure matrix. If the brightness of a pixel in the correction image is greater than the local mean, the ratio
Ilocal current mean/Icurrent
results in a value of less than 1. If the brightness value of a pixel in the correction image is darker than the local mean, this results in a value of greater than 1. The structure matrix generated in step g and illustrated in a graph in
Very good results are obtained by the method described above of averaging the brightness distribution over adjacent regions, for example over a block of 36 or 49 pixels; however, it is very computationally intensive. In order to simplify this, it is possible to use only local values from 20 pixels of the respectively current video line for both the vignetting and the structure matrix. This results in a simplified vignetting matrix as illustrated in
As per the flowchart illustrated in
Particularly for the purpose of editing relatively large amounts of data, for example for the purpose of generating correction matrices for a number of lens apertures of different types of cine lenses, the calibration images may, for external data processing, be output as video signals at the image output of the image electronics 5 and transmitted from there to an external computer, for example as a video image via a frame-grabber card. The correction matrices are calculated in the external computer and all correction matrices, or the respectively required ones, are output to the image processing unit 7 via the data interface 70, for example an Ethernet, USB or similar interface, in which image processing unit they are stored in the vignetting-matrix memory 74 and structure-matrix memory 75 for pixel-by-pixel comparison with the real recorded images.
Correcting the actual recorded images from the digital video camera must be implemented in a real-time capable system, for example by means of a field programmable gate array. The individual pixels in the recorded image 71 are first of all multiplied by the pixel output by the vignetting-matrix memory 74 that is stored at the same address. This is followed by multiplication by the pixel with the same address output by the structure-matrix memory 75. By multiplying each pixel in a recorded image by the same pixels in the calibration image, points that are too dark as a result of the pixel graininess or vignetting are multiplied by a factor of greater than 1 and hence these are brightened, whilst points that are too bright as a result of the pixel graininess or vignetting are multiplied by a factor of less than 1 and hence these are darkened, and so, overall, the recorded image is freed from the influence of the grain structure and vignetting.
Use can be made of two different methods for taking account of the influence of the lens aperture in the cine lens 1.
In a first method, a plurality of correction images are established for different diaphragms and these are then reapplied in the case of the corresponding diaphragms in the recording mode.
In a second method, use is made of the two additional correction factors 76, 77, which act on the two correction matrices. To this end, provision is made for the first multiplier 781 and the third multiplier 783, with the correction factors 76, 77 being dependent on the respective lens aperture of the cine lens 1.
Number | Date | Country | Kind |
---|---|---|---|
10 2009 002 393.3 | Apr 2009 | DE | national |
This application is a National Phase Patent Application of International Patent Application Number PCT/EP2010/054882, filed on Apr. 14, 2010, which claims priority of German Patent Application Number 10 2009 002 393.3, filed on Apr. 15, 2009.
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/EP2010/054882 | 4/14/2010 | WO | 00 | 10/13/2011 |