EP2478705A1

EP2478705A1 - Method and device for generating partial views and/or a stereoscopic image master from a 2d-view for stereoscopic playback

Info

Publication number: EP2478705A1
Application number: EP10749808A
Authority: EP
Inventors: Michael Steglich
Original assignee: NATURAL VIEW SYSTEMS GmbH
Current assignee: NATURAL VIEW SYSTEMS GmbH
Priority date: 2009-09-15
Filing date: 2010-08-27
Publication date: 2012-07-25
Also published as: WO2011032642A1; CN102612837B; RU2012114692A; US8693767B2; CN102612837A; AU2010294914B2; US20120170837A1; DE102009041328A1; RU2541928C2; AU2010294914A1

Abstract

The invention relates to a method (1) and to a device (31, 51) for generating image data, in particular a three-dimensional image master, for stereoscopic playback via an autostereoscopic playback device. The method (1) for generating image data for playback that evokes a three-dimensional visual impression in a human observer comprises the following steps: receiving and/or importing of, preferably colour, 2D image data of an original view (2), deriving a plurality of further partial views (7-1 to 7-4) that represent the information shown in the original view (2) from different viewing angles, wherein for each pixel of the 2D image data of the original view (2), colour values of at least two primary colours of a colour system are determined, and a control value (8, 23-k) dependent on the at least two determined colour values is automatically assigned to each of the pixels (21-k). Each control value (8, 23-k) represents disparity information, and while deriving the plurality of partial views (7-1 to 7-4), which represents different viewing directions of an original scene represented by the 2D image data of the original view (2), the individual pixels (21-k) of the original image are shifted pixel-wise relative to one another according to a displacement vector pre-set for the respective view and corresponding to the control value (8, 23-k) assigned to the individual pixel, and the pixels of the respective view are derived on the basis of the shifted pixels (25-I). The device according to the invention (31, 51) preferably implements the method (1) according to the invention by means of an application-specific integrated circuit, so that, for example, for each view of an image sequence a plurality of partial views (7-1 to 7-4) is automatically calculated at the clock rate of the image playback of an image sequence (i.e. in video replay real time) and, if necessary, the respective partial views can be combined to form a stereoscopic image master for an autostereoscopic playback device (9).

Description

Method and device for generating partial views and / or a spatial image template from a 2D view for stereoscopic reproduction

The invention relates to a method and a device for generating partial views and optionally a spatial image template for a stereoscopic reproduction via an autostereoscopic reproduction device. For this purpose, from a two-dimensional image, which is an original view and images an origin scene from a viewing angle, several further views, which are referred to as partial views, are derived, which represent the original scene from other viewing angles. From these further views, as a rule, possibly incorporating the original view, which likewise represents a partial view, a spatial image template is created which generates a three-dimensional visual impression of the original scene in a reproduction via the autostereoscopic display device in a human observer.

Illustrations of the real, three-dimensional world, as well as mathematically or artistically generated three-dimensional scenes and objects have always been depicted on two-dimensional surfaces and presented in the form of a projection. The depth information is lost in this case, but man already learns in early childhood to reconstruct part of this information based on his experience and object knowledge.

In order to obtain the depth information, various methods for detecting and subsequently displaying are known in the prior art. For example, the real three-dimensional world can be detected stereoscopically. For this purpose, the same scene is detected by means of at least two detection devices from different viewing directions or viewing angles. The detection must be done simultaneously for moving objects in the scene. The detection can be done, for example, photographically by film or digitally by means of a video camera, a CCD, an electronic camera or the like. The different views can also be generated computationally or artificially, whereby the scenes and objects are detected from at least two directions and the resulting images are subsequently made visible to the left and right human eye in a suitable manner. By recombining the images in the visual perceptual apparatus of the human brain, a true three-dimensional image impression can be generated in the observer.

There are various technical solutions for the reproduction of stereoscopically recorded or calculated images. Classic are stereo slide image viewer for each to call a viewer. Also known are various methods for delimiting the right and left image by means of coloring and subsequent separation of the shares by using differently colored lenses. A well-known method uses the polarity property of the light, whereby the different images are offered to the eyes of the observer simultaneously but with different polarization directions and separated by corresponding polarization filters (goggles) in front of the right and left eyes. Electronically controlled shutter glasses are also used for the stereoscopic display, in which the viewer is offered the right and left images alternately by means of technical devices and at a sufficiently high speed, the respective right or left spectacle lenses being alternately switched to black at the same frequency. With sufficient speed, the human visual apparatus is no longer able to perceive the individual images separately, so that the result is again the impression of a three-dimensional overall image, although the two eyes are supplied with different image information (views) at different times. The methods operating on the basis of the various spectacles mentioned have the advantage that even a number of spectacles-equipped observers can at the same time see the displayed image information in an adequate manner, the use of spectacles being subjectively often perceived as a disadvantage.

For this reason, various autostereoscopic visualization systems have been developed that are state of the art today. In these display devices or playback devices, multiple images from different recording positions are simultaneously superimposed and optically separated in a device placed in front of the imaging plane, so that the eyes of one observer can each see two different images. In the case of optoelectronic displays, this is achieved by limiting or modifying the emission direction for individual pixels and thus, in a restricted viewing space, different partial images from different directions can be seen. If the angle of emission of two adjacent images at a certain distance corresponds to the distance between the eyes, a three-dimensional image can thus be generated by the observer even without glasses. Barriers, lenticular sheets and microlens arrays are known as suitable devices for separating the emission direction.

From the application US2007 / 0046899 by Shuji Moro et al. is a novel display technology is known in which a low-loss radiation of the individual images takes place using a reflection module. The autostereoscopic display is usually associated with a loss of pixel resolution and in many cases (eg, the use of barriers) with a loss of image brightness due to the required, simultaneous display of multiple superimposed images. Some autostereoscopic solutions also have the disadvantage that in each case only a single observer can perceive the stereoscopic effect, with elaborate tracking systems are used, with the help of the viewer's own motion or eye movement are detected in space before the image plane and then compensated.

The aforementioned methods for generating three-dimensional image impressions in humans presuppose that in each case there are at least two different views of the represented scene or of the objects which have been acquired or calculated from different directions. For capturing this type of artwork, multi-camera systems and corresponding image storage and transfer means are required. For virtually generated scenes, this task is usually relatively easy to solve by rendering two views of the scenes with different virtual camera positions.

However, there are already huge amounts of monocular recorded image and film material in a variety of digital and analog formats, which belong to the fund of human culture. Almost the entire studio and recording technique for image information as well as transmission channels and visualization technology are currently designed for monocular recorded image information. A doubling of the bandwidths, as would be necessary for the transition to stereoscopic image information, is partially technically impossible and economically hardly feasible.

As stated, providing multiple views for virtually generated scenes through multiple rendering is usually not a problem: since the required resolution for the stereo pairs is lower, in this case, there is no much greater computational overhead. However, this only applies to the reproduction on a particular reproduction device and a predetermined optimal viewing distance of an autostereoscopic display. The problem with regard to the required increased bandwidth in the transmission remains, however, if a reproduction on any autostereoscopic playback devices should be done, since then the fields are to be transmitted at full resolution. The situation is different for recordings of the real world (film, television) and for the existing fund of film and video material, including traditionally produced animated films. Even for older, virtually generated material (computer games, presentations, etc.) is a recalculation not unproblematic in stereoscopic form.

In addition to the aforementioned difficulties and problems with recording and transmission technology, there are further obstacles in the way of the widespread use of stereoscopic methods. For example, modern and video production has made optical and computational effects such as crossfades, overlays, trick techniques and the mixture of reality and virtual reality indispensable. These effects are due in no small part to the mix of 2D and 3D imagery. Most of the methods and tools could no longer be used in the case of stereoscopic reproduction, or could only be used to a very limited extent.

Furthermore, the experimental use of stereoscopy has shown that the viewing of stereoscopically recorded and autostereoscopically reproduced image information can cause problems with the visual perception in the viewer. These problems are caused by the fact that in stereoscopic recordings the recipient decides via the focus of the recording optics used which image plane is to be viewed or viewed, but the viewer is accustomed to his experience, his eye independently and at his own discretion to an individual To focus objects or in a viewing distance. Due to this dilemma, the resulting visual impression of the viewer does not correspond to the natural perception despite the stereoscopically correct representation. An occurrence of physiological, psychological and / or medical problems, for example in the Akkodomation of the eyes, is known with prolonged consideration of such image sequences.

Due to the technical, economic and other problems mentioned, efforts have been made for some time to compile traditionally monocular image information in such a way that it can be visualized using generically generated, additional depth information via stereoscopic display technology. In particular, because of the lack of real depth information, these approaches are based on upstream image analysis and image segmentation operations, which can be performed either interactively by an operator or by combined image processing techniques. In DE 103 48 618 A1, a method is described in which by means of upstream interpretation and segmentation of an existing image by analysis of the structure resolution or the color components predominantly present in certain areas, a segmentation into individual, contiguous areas takes place, which then more or less arbitrary different depth values can be assigned manually or interactively via an operator. The basis for the segmentation are empirical rules. In exceptional cases, segmentation can be used to automatically derive content and assign depth values to the content. However, content-related requirements are necessary for this.

In the method described, the determined segments are assigned to individual depth levels; a continuous, continuous depth map for the output image can not be generated. Further, highlighting or dropping in the background of individual image segments is explicitly intended to interprete the image content, i. bound to the detected objects in the scene. Thus, DE 103 48 618 A1 expressly emphasizes the content-dependent segmentation method. Automated recognition of this image content is possible, if at all, only in special cases.

An automated generation of multiple views of a two-dimensional recorded original scene and / or derived therefrom spatial image template is not possible with the methods of the prior art.

The object of the invention is therefore to provide a method and an apparatus which automatically generate from a two-dimensional (2D) view of an originating scene further views of the originating scene, representing the scene from different viewing angles or viewing directions and from this, if necessary, combine a spatial image template that at a reproduction via a stereoscopic or autostereoscopic display device as realistic as possible three-dimensional visual impression is caused in a viewer. The method and apparatus are preferably designed to be able to automatically process individual two-dimensional views of a sequence in real time, ie. without requiring human intervention during processing.

The invention is based on the finding that the receptors for different color stimuli on the human retina are not evenly distributed. Therefore, different shades have a different influence on the spatial vision of humans. A method is proposed for generating image data for a reproduction that provides a three-dimensional visual impression in a human Obtaining viewer, comprising the steps of receiving and / or reading in 2D image data of an original view, deriving from several other views, representing the information shown in the original view, which are referred to as the original scene, from different viewing angles. These views, which represent the originating scene from different viewing directions or viewing angles, are also referred to below as partial views. According to the invention, color values of at least two primary colors of a color system are determined for each pixel of the 2D image data of the original view, each of the pixels is automatically assigned a control value as a function of the at least two determined color values, each of the control values representing disparity information, and during deriving of the plurality of views representing different viewing directions of an original scene represented by the original image 2D image data, the individual pixels of the original view are shifted pixel-by-pixel relative to each other in accordance with a shift vector given to the respective partial view and the control value associated with the single pixel, and from the shifted pixels Origin view the pixels of the respective derived partial view are derived. It is crucial that the control values that automatically derive the disparity of individual pixels in the different views together with displacement vectors that define a viewing angle change between the individual views are dependent only on the individual color values that are assigned to the individual pixels of the original view. The multiple partial views of the original scene shown in the original view can thus be determined completely automated without human intervention. Disparity or disparity information is understood here as a measure which indicates how large an offset of the imaging positions of the same imaged object is in different partial views that represent the same original scene from different viewing directions. Thus, there is provided an apparatus for generating image data for reproduction via a stereoscopic or autostereoscopic display apparatus, the rendering of which via the display apparatus produces a three-dimensional visual impression on a human observer, the apparatus comprising: an interface for receiving 2D digital image data Origin view, and a view determination unit for deriving a plurality of derived subviews that respectively represent the originating scene represented in the original view from a different viewing angle. According to the invention, the view determination unit comprises a color value determination unit which determines color values for at least two primary colors of a color system for each of the pixels of the original view; an allocation unit, the each of the pixels depending on the at least two determined color values one Assigns control value, and a shift unit, which shifts for each of the other views every single pixel of the original view depending on the associated control value and a predetermined displacement for the corresponding view vector and derives the pixels of the respective partial view based on the shifted pixels. This means that the visibility and, if appropriate, the location of the visibility and, if appropriate, the location of the visibility and the color value in the respective partial view are determined with the inclusion of the adjacent pixels on the basis of the associated control value and the displacement vector. Advantage of the invention is that in a simple way, several derived partial views can be created without the intervention of a person is necessary or a content of the original scene shown in the original view is required. Depending on the choice of viewing directions or viewing angles for which the individual partial views are derived, the original view also represents a partial view to which the displacement vector 0 is to be assigned. This is the case when one of the partial views is assigned the original viewing direction (or the original viewing angle or the original viewing position). Even if one of the partial views is identical to the original view, this partial view is also considered as a derived partial view. This means that the derived partial views may include but are not necessarily included in the original view. In many cases, all partial views that are derived differ from the original view.

The representation of the acquired image data is usually via autostereoscopic playback devices. These are usually supplied with a so-called spatial image template. Such can be created by merging the multiple derived partial views, which may include the original view. A preferred development thus provides for combining the derived partial views into a spatial image template. A corresponding device thus comprises an assembly unit for creating the spatial image template. The way in which the several derived partial views, which optionally comprise the original view, are combined to form a spatial image template, depends on the reproduction apparatus to be used. The necessary information is usually given by the manufacturer of the playback device.

In one embodiment, the control values for the 2D image data of the original view are determined at a central location, for example a transmitting station, and assigned to it. Thus, in the transmitting station, a method is carried out comprising the steps of: receiving and / or reading in 2D image data of an original view, inferring and Assigning a control value to each pixel of the original view; wherein color values of at least two primary colors of a color system are determined for each pixel of the original image 2D image data and each of the pixels is automatically assigned a control value dependent on the at least two determined color values, each of the control values representing disparity information.

Subsequently, the 2D image data are transmitted together with the control values to one or more receiving stations, which is indicated by an arrow. It is understood that the transmission can be coded or uncoded depending on the transmission link. When coding a coding can be selected, which makes it possible to detect transmission errors and possibly even correct.

The following steps are carried out in the receiving stations: receiving and / or

Reading in 2D image data of an original view and control values associated with pixels of the 2D image data, each of the control values representing disparity information and being determined for the associated pixel of the original image 2D image data from color values of at least two primary colors of a color system or automatically determinable therefrom Deriving a plurality of further partial views representing the information shown in the original view from different viewing angles;

wherein, in deriving the plurality of partial views representing different viewing directions of an original scene represented by the 2D image data of the original view, the individual pixels of the original view are pixel-wise corresponding to a displacement vector and the individual given to each view

Pixel associated control value are shifted relative to each other and based on the shifted pixels, the pixels of the respective partial view are derived.

An apparatus for image data processing for generating image data for reproduction via a reproduction apparatus, in the reproduction of which via the reproduction apparatus a three-dimensional visual impression is produced by a human observer, comprises in such a receiving station: an interface for receiving 2D digital image data of an original view and of control values, wherein each pixel is assigned exactly one control value and each of the control values represents disparity information and is determined for the associated pixel of the original image 2D image data from color values of at least two primary colors of a color system of that assigned pixel or can be automatically determined from these color values; a view determination unit for deriving a plurality of partial views representing the original scene represented in the original view from different viewing angles; and a Summary unit for creating a room image template from the several

Partial views, wherein the view determination unit comprises a shift unit which shifts for each of the partial views each of the individual pixels of the original view depending on the associated control value and a given for the corresponding partial view displacement vector and derived from the shifted pixels, the pixels of the respective partial view.

A transmitting station comprises an image data processing device for processing image data for playback via a playback device, the playback of which via the playback device produces a three-dimensional visual impression on a human observer, the device comprising: an interface to

Receiving digital 2D image data of an original view;

a color value determination unit which determines color values for at least two primary colors of a color system for each of the pixels of the original view, as well as

an allocation unit which assigns a control value to each of the pixels as a function of the at least two color values determined. Preferably, the 2D image data and the associated control values can be output or provided via the interface for receiving the 2D image data or a further interface, for example a transmission interface.

The control values assigned to the individual pixels are calculated in a preferred embodiment. In a development of a device according to the invention, a calculation unit is provided for this purpose. Here, a function is used that depends on the at least two specific color values as variables.

The function for the calculation of the control values as a function of the determined color values is in a preferred embodiment a function which has an extreme value and is continuous or discrete continuous, but need not be continuously differentiable.

In addition to the dependence on the at least two determined color values, the function may still depend on parameter values determined by the reproducing apparatus with which the generated image data is to be reproduced. Moreover, in some embodiments, a given optimal viewing distance has an influence on the calculation of the control values.

A color system is a system which makes it possible to characterize a color using the color values of individual primary colors. A well-known color system for Characterization of colors is, for example, the RGB system (red-green-blue system), in which the individual colors can be generated via a color addition of a red color component, a green color component and a blue color component of the primary colors red, green and blue. When determining color values for a color, the proportion of the individual base color with which it participates in the color addition is determined in each case. For example, in the RGB color system, color values between 0 and 100 are determined for each primary color. The color values then indicate a percentage of an intensity with which the respective base color enters into the color addition. If, for example, all colors red, green and blue at 100% intensity enter color addition in an RGB color system, the result is a white color for the human viewer. If red and green with 100% intensity and blue with 0% intensity are included in the color addition, the result is, for example, a yellow hue.

In particular, if several original views of a sequence of images are to be processed for a three-dimensional rendering, which are then to be output on the same reproduction apparatus and viewed at the same distance thereto, it is advantageous if the control values, which then depend exclusively on the at least two color values, only once for all possible color value combinations and stored in a control value memory. When the control values are assigned to the individual pixels, the control values can then be read from the control value memory, depending on the determined at least two color values.

In another embodiment, the control values are each individually calculated currently for each pixel based on the control value function. This is useful if image data is to be processed for only one source view, or if successively image data from different source views are to be processed, but which should be viewed on different display devices or at different distances to the playback device (s).

In order to use rendering devices that not only create a spatial effect from a viewing position, it is necessary that more than two views be derived from the original view. In a preferred embodiment n partial views are derived, where ns 2 applies and the shift vector V _{m associated with} the partial view is given the partial view m with m = 1,... N by: V _m = a ^* (-1 + 2 ^* (m -1) / (n-1)), where a is a constant, and preferably: a = 1. In this embodiment, an equal number of partial views are generated, which are opposite to an original viewing position from which the original view is taken Representing the left-shifted viewing position how views are generated that represent a right-shifted viewing position relative to the viewing position selected for the origin view. If n is odd, the original view represents one of the partial views, namely the partial view m = (n + 1) / 2, which is preferably used for calculating a spatial image template. It is advantageous that the viewing positions are each shifted by the same distance. This is ensured in the described preferred embodiment on the choice of the displacement vectors.

The shifting of the individual pixels of the original view preferably takes place in such a way that the respective pixels are shifted in accordance with a disparity vector, the disparity vector for each pixel being calculated as a product of the control value associated with the pixel and the shift vector of the corresponding view. The disparity vector is thus for each pixel of the corresponding control value weighted shift vector of the corresponding view. The pixels of the source view are processed one row at a time, one at a time, in the direction pointed to by the respective displacement vector associated with the partial view being derived. The pixels of the original view, which are moved to positions which (with regard to the processing direction) remain behind a position to which a pixel of the line has already been moved, are ignored. This is due to the fact that pixels which have a lower control value than their neighboring pixel, which is arranged with respect to the displacement direction "in front of them", are not visible in the generated derived partial view, which represents a changed viewing angle from the original view. The disparity expressed by the control value thus also determines a visibility of the corresponding pixel in relation to the disparities (control values) of the neighboring pixels (not necessarily the immediately adjacent pixels). Pixels that will not be visible in one of the partial views need not be moved accordingly for this partial view. This is equivalent to disregarding the corresponding pixel of the original view. The pixels of the derived view, the partial view, are then derived from the shifted pixels.

If the control values are integral and the magnitude of the displacement vector V is also an integer value, the pixels assigned to a matrix, which have vertical columns and horizontal rows, are automatically shifted to a new matrix position during the displacement process, assuming that the distance between the individual matrix positions an integer multiple of the pixel spacing and the control value range and the value range of the lengths of the displacement vectors are selected accordingly. Since a resulting offset of the individual pixels of the original view is exclusively determined by the pixel-related control values and the partial view-related displacement vectors, defects may arise in the pixel structure or matrix assigned to the shifted pixels if disparity vectors of the pixels adjacent in the original view differ from one another. These are partly not shifted or not to be considered pixels of the original view to assign, as already explained above. However, defects also arise when the control value of the pixels increases in the processing direction. A subsequent pixel is thus shifted by a larger amount than the previous pixel, so that between these one or more defects may arise. Irrespective of their origin, the defects can be filled up either by the color values of the shifted pixel, which is arranged with respect to the displacement direction in front of the respective defect or pixel of the pixel structure or matrix, ie is less or not shifted than the pixel that far it is postponed that one or more other pixels having a lower control value are or are considered to be non-visible and therefore remain or remain undisclosed and thus cause or cause the one or more defects.

In another embodiment, it is provided that the resulting defects are provided in a line with color values, which are determined by an interpolation from the color values in the derivation of the view to be considered displaced pixels that surround or limit the defects after the shift. If the amounts of the disparity vectors formed from the displacement vectors and the control values are not integral, then it is necessary for all or most of the pixels of the partial view matrix to determine the color values for the individual matrix positions either on the basis of the color values of the pixel closest to the respective line or to calculate the color values by interpolation from the color values of the shifted pixels to be considered. The interpolation may be performed linearly or performed following another predetermined function. It is also possible to take into account color values of shifted pixels that are neighbors next to, or next to, etc. neighbors. It should be noted that, of course, each of the pixels of the partial views is assigned a complete color information which comprises color values with respect to all primary colors of the color system. In any case, however, the computational effort for generating the individual partial views to be derived from an original view of a sequence can be kept so low that the calculation of the multiple partial views can be done in "video / TV real-time", ie after a possibly necessary start-up phase, the multiple partial views in FIG It should be noted that both the individual partial views can be derived in parallel and the individual rows can be processed line by line in parallel independently of each other.This parallelization makes it possible, even with high original viewing rates to calculate the required derived partial views for determining a spatial image template with the clock rate of the image sequence in a reproduction of an image sequence.

Since the received 2D image data in the derivation of the partial views and the control values are usually accessed multiple times, in some embodiments these are stored in a first memory area. Likewise, it is provided in some embodiments to deposit the determined partial views or at least parts thereof in second storage areas.

The individual units of the device for generating the required image data are preferably realized via a software-controlled computer unit. However, it is also possible to integrate the individual units in one or more FPGA units. Yet another embodiment provides that an ASIC circuit converts the individual units, wherein the image data of the original view and, if appropriate, parameter values which depend on the device or viewing distance can be read in via data lines and the generated image data are output via data lines. In this case, embodiments can be implemented in which the first memory area and the second memory areas can be integrated into the ASIC circuit or can also be located in a memory which is coupled to the ASIC circuit. The same applies to the control value memory, which however is preferably implemented in the ASIC circuit.

The invention will be explained in more detail below with reference to a preferred embodiment with reference to a drawing. Hereby show:

Fig. 1 is a schematic flow of the processing of image data of a 2D image to

Processing and presentation as a 3D image; Fig. 2 is a schematic representation for explaining the image point wise

Displacement of pixels to produce the derived partial views;

3 is a schematic representation of a device for processing 2D

Image data for a three-dimensional representation;

Fig. 4 shows a further embodiment of a schematic representation of a

Apparatus for processing 2D image data for a three-dimensional representation;

Fig. 5 is a schematic representation to illustrate an optimized

Processing of 2D image data for three-dimensional representation;

FIGS. 6a to 6c show exemplary assignment matrices for the combination of partial views into a spatial image template; and

7 shows a schematic representation of a method in which control values are assigned to the 2D image data in a transmitting station and the partial views are derived at the receiving stations from the received 2D image data of the original view together with the received control data and combined to form room image templates.

FIG. 1 schematically shows a method 1 of processing 2D image data of an original view 2. First, the 2D image data are received, recorded or read in the form of an image file, for example. The image file of the 2D image data here comprises digitized pixels of the original view 2, which represents an original scene. The received 2D image data is preferably colored 2D image data. For each pixel of the original view 2 here is a color information available. For example, a total of 24 bits are available for encoding the color, with 8 bits each being used for coding a color value in an RGB color system. However, the received 2D image data may have gray scale encoding, with each pixel being associated with a gray level.

The received 2D image data may be the image data of a single image, which originates, for example, from a data carrier 3 or may be a view of an image sequence 4, for example from a receiver / playback device 5 provided. The receiving / reproducing apparatus 5 may be, for example, a television receiver, a video recorder, a DVD player or other player.

The 2D image data is processed in a processing stage, which may also be referred to as a view determination unit 6, to derive therefrom several partial views 7-1, 7-2, 7-3, 7-4. The plurality of derived partial views 7-1 to 7-4 each have the same number of pixels as the original view represented by the 2D image data. If a change in format, resolution, etc., is desired in the intended 3D reproduction in comparison with the original view 2, then the original view is adapted accordingly prior to processing, so that this condition applies to this adapted original view and the partial views derived therefrom.

Each of the derived partial views 7-1 to 7-4 represents the original scene represented by the original view 2 from another viewing position or another viewing angle or another viewing direction. The viewer determining unit 6 first makes at least two color values of Basic colors of a color system determined. In the embodiment shown here, in which the 2D image data of the original view are provided in the form of an RGB-coded image file, the corresponding color values, for example for the primary colors red and blue of each pixel, can be read out of the 2D image file , If the 2D image file is provided in a different color system or in gray scale coded values, the color values may need to be calculated. Preferably, the color values are converted into the RGB color system. For a conversion from a greyscale coding into an RGB coding or a conversion from one color system to another color system, there are calculation rules known in the prior art.

The pixels of the original view 2 are individually assigned control values 8 on the basis of the determined color values. The control value 8 in each case represents a disparity information which, in the later processing sequence, determines how great the disparity for the respective pixel in the individual derived partial views 7-1 to 7-4 with respect to the original view 2 or between the individual derived partial views 7-1 to 7-4 will be. If one takes a real scene of two slightly shifted viewing positions, one finds that pixels that represent the same object and have a large disparity between the two shots often "obscure" pixels of objects that would have less disparity Thus, the control values 8 also regulate the visibility of individual pixels of the Origin view 2 in the derived partial views 7-1 to 7-4, as will be explained below.

The individual pixels of the 2D image file can be assigned to a pixel matrix having vertical columns and horizontal rows. In order to generate the individual derived partial views 7-1 to 7-4, the 2D image data of the original view 2 are processed plural times line by line to generate respective pixel lines for the plurality of derived views 7-1 to 7-4. Each line is processed as often as derived partial views are generated. (If one of the derived partial views is identical to the original view due to the specifications with regard to the viewing directions assigned to the partial views, processing of the image row can, of course, be saved.) In the illustrated embodiment, in which four derived partial views 7-1 to 7-4 are generated or are derived, each line of the original view 2 is thus processed four times. This is assumed in each case from the image line of the original view 2. Each of the partial views 7-1 to 7-4 to be derived is assigned a displacement vector V _m , where m indicates an index of the respective view, ie m = 1 n and n indicates the number of derived views. The derived partial views

7-1 to 7-4 respectively represent the original scene represented by the origin view 2 from different viewing positions. These viewing positions are or are calculated in equal numbers equidistantly symmetrically right and left to the viewing position from which the original view is taken. The displacement of the viewing position relative to the viewing position of the original view is represented in each case via a displacement vector V _m , which is assigned to the respective partial view m. With an equally spaced symmetrical distribution of the viewing positions for the partial views to be derived, a calculation rule for the displacement vector V _m = a ^* (-1 + 2 ^* (m-1) / (n-1)), where a is a constant and preferably results equal to 1 is selected. The displacement vector V _{m associated} with the respective partial view m is weighted or scaled with the control value associated with the individual pixel in order to determine a disparity vector for each individual pixel of the original view 2 in the line-by-line processing. For each pixel of the original view 2, as many disparity vectors are thus determined as partial views are derived. While the control values each have the same sign, both positive and negative shift vectors occur. The sign of the displacement vector is associated with a shift direction. Within a pixel row, a positive sign is associated with a direction pointing from left to right and a negative sign corresponding to a direction pointing from right to left. The individual pixels of a row are corresponding to that direction processed, which is predetermined by the direction of the displacement vector for the corresponding partial view. If the shift vector is positive, the pixels of a line are processed from left to right and shifted from left to right according to the respective disparity vector. It should be noted that those pixels whose disparity vector has a length which is shorter than the length of the previously processed pixel, due to the disparity in the corresponding partial view that is calculated, are not visible and therefore not moved or for the Derivation of the pixels of the corresponding partial view are not taken into account. This also means that pixels whose disparity vector would assign them a new position, which is already occupied by a previously shifted pixel with a larger disparity vector, are not visible and therefore will not be moved or taken into account. Since this can not occur only with respect to the immediately pre-processed pixel, it is generally understood that a pixel that would be shifted with the disparity vector to a position that remains behind a position with respect to the processing direction to which a pixel of the line has previously been moved , is not moved or is not taken into account in the derivation of the pixels for the corresponding partial view.

In one embodiment, the amounts of the disparity vectors are rounded to integers so that the shifted pixels can be associated with pixels of a matrix-like raster of the partial view to be derived. The shifted pixels thus represent directly the derived pixels of the partial view.

Since the disparity vectors originally (not necessarily immediately) of adjacent pixels of the original view may differ greatly from each other, as explained above, gaps between the shifted pixels taken into account in the derivation of the pixels of the partial view will appear in the one with the corresponding partial view associated matrix-like raster. In one embodiment, these imperfections are filled with the color values which correspond to the pixel previously shifted and to be taken into account with respect to the execution direction of the line. In another embodiment, the color values of the defects are calculated by interpolation of the adjacent color values to be considered.

Instead of rounding the lengths of the disparity vectors to whole numbers, the color values for the individual pixels can also be calculated from the pixels to be considered, whose positions do not coincide exactly with the pixel positions of the matrix-like grid of the partial view, by a subpixel-accurate interpolation. The pixels of a line can each independently of the pixels in the adjacent lines are determined. Likewise, the pixels of the different derived partial views 7-1 to 7-4 to be generated at the same time can be independently derived from the pixels of the image data representing the original view 2. As a result, it is thus possible to parallelize the calculation of the individual derived partial views 7-1 to 7-4 in terms of time. It is thus possible, even with high-resolution image formats with a high frame rate, to calculate the multiple derived partial views 7-1 to 7-4 in real time, ie with the frame rate.

The individual determined lines of pixels of the derived partial views 7-1 to 7-4 are stored in the embodiment described here in each case in corresponding image memory. In order to ensure reproduction via a stereoscopic reproduction apparatus 9, the plurality of derived views 7-1 to 7-4 are assembled into a spatial image original 10 in a so-called recombination stage 11, which is output to the stereoscopic reproduction apparatus 9.

It is again noted that the original view prior to processing, i. a derivation of the plurality of views, is recalculated to the playback format, if the resolution of the playback device differs from a resolution of the original view. This is not shown in Fig. 1, but may optionally be provided.

The control values 8 are calculated in a preferred embodiment according to a function that depends on the blue and red color value component of the pixels. The function has an extremum which occurs at a red value R ^lim and at a blue value B ^{l, m} . The red and blue color values R ^{, im} , B ^lim at which the maximum of the control value function occurs are referred to as the red color threshold and the blue color threshold. First, the absolute values of the differences of the respective color value and the corresponding color limit value are determined. The control value is determined by the larger of these absolute amounts divided by a normalization constant. If the absolute value of the color value of the red color portion of a pixel of the original view 2 reduced by the amount of the red color threshold is greater than the absolute value of the color value of the blue color fraction reduced by the amount of the blue color threshold, the control value is the absolute value of the difference of the red color value and If, on the other hand, the absolute value of the color value of the blue color component reduced by the amount of the blue color limit is greater than the absolute value of the color value of the red color component reduced by the amount of the red color limit, the control value becomes the absolute value of the difference of the blue color value and the blue one Color limit determined. The absolute values of the corresponding differences are each normalized by a common factor.

Generalized:

Be ^ and f ₂ color values of a pixel with respect to two primary colors then the control value S fa) is calculated according to the following formula:

S (fi, f ₂ ) = (| f ₂ -f ₂ ^lim | / const.), Otherwise.,

where const. stands for a normalization constant and ^m and f ₂ ^{hm are} color threshold values of the primary colors at which the control value function S (fi, f ₂ ) has an extremum. The function is steadily above the area of the color values and f ₂ . If the color system is suitably selected, the control values depend on exactly two color values.

Since the control values for a number of image data sets, which represent different original views of an image sequence, remain unchanged, the control values can be stored in and output from a control value memory and only have to be calculated once for the corresponding image sequence. If the reproducing apparatus used and the desired optimum viewing distance remain constant for different image sequences, the control values need only be calculated once.

It will be explained with reference to FIG. 2 how the individual pixels 21-k (k = 1, 2,...) Of the original view of a line 22 are shifted in order to obtain the pixels 23-I (I = 1, 2,. .) To generate a line 24 of a partial view m. The first eight pixels 21-1 to 21-8 have already been shifted in the processing steps S1 to S8 to the shifted pixels 23-1 to 23-8. For the processing steps S9 to S18 is shown in each case how the individual pixels are moved in the individual steps or not moved and not taken into account. The individual pixels 21-k are shown as boxes, their color, which is defined by the individual color values, being schematically indicated by a hatching. The individual pixels 21-k of the image line 22 of the original view are assigned the individual control values 23. In method step S9, the pixel 21-9 to which the control value with the value 6 is assigned should be shifted. The pixels 21-1 to 21-8, to each of which a control value 0 is assigned, have already been shifted. The pixel 21-9 to be displaced is shifted to the right by six pixel positions in the image line 24 of the view to be generated and thus represents a partial view pixel 25-15. Between the pixel 25-15 of the partial view generated in step S9 and the last one during translation of the original pixel Spaces 25-9 to 25-14 have been created at 21-8 designated pixels 25-8 of the partial view to be generated. These are filled with the color value corresponding to the pixel which was last set before the currently shifted pixel 25-15, namely the pixel 25-8. This is shown schematically in the bracketed image line 24 'in the processing step S9'. In the subsequent processing steps S10 and S11, the pixels of the original view 21-10 and 21-11, to which a respective control value with the value 6 is assigned, are shifted into the pixels 25-16 and 25-17. In the subsequent processing step S12, a control value with the value 2 is assigned to the pixel 21-12. If you were to move this pixel 21-12 by two pixel points, so an already specified pixel, namely the pixel 25-14 would be overwritten. Since points which have a lower control value and thus a lower disparity in the shifted views than the previously processed pixels which have a control value of the value 6, this pixel 21-12 is virtually hidden in the corresponding partial view due to the disparity. Thus, this pixel 21-12 is not shifted or taken into account in the definition of the color values for the pixels of the partial view to be generated. The same occurs in the subsequent processing steps S13 to S17, in which the pixels 21-13 to 21-17, to each of which a control value 0 is assigned, would be moved to a previously defined pixel of the partial view to be generated. In the processing step S18, however, the pixel 21-18, to which a control value of 0 is assigned, is used and shifted by 0 pixel locations, thus defining the pixel 25-18 of the line 24 to be generated of the corresponding partial view.

As an alternative to filling the imperfections with color values which correspond to the previously previously shifted pixel, as shown schematically in method step 9 or 9 ', the imperfections can also be obtained via an interpolation of the color values, here for example pixels 25-8 and 25-8. 15, done. In the illustrated example, it is assumed that the control values or more precisely the amounts of the disparity vectors, which are generated by a multiplication of the control value by the corresponding shift vector, are rounded to integer values. In other embodiments, the color values can be interpolated since the pixels of the original view are then not shifted by exact image locations or pixel spacings.

It is noted that the displacement of the pixels with respect to a matrix-type raster directly results in the pixels of the derived partial view only in such cases in which the disparity vectors used during the displacement are rounded to integers in length (if they do not already have an integer length) and the whole Number shifts are assigned around whole matrix locations. In other cases, the shifted pixels form a pixel structure which, even ignoring the possibly existing "defects" or the pixels which are not to be considered for the displacement, does not necessarily have to have a regular matrix structure , on the basis of which the actual pixel matrix of the partial view is determined. This means that the pixels of the pixel matrix of the partial view are determined on the basis of the color values of the pixel structure (intermediate matrix).

FIG. 3 schematically shows an apparatus 31 for image data processing for generating image data for a 3D reproduction. This is designed to use 2D color image data for generating a plurality of partial views thereof and, if necessary, to combine these partial views into a spatial image template. Via an interface 32, the image data of the original view and, if appropriate, parameters via the reproduction device and / or an optimal viewing distance to the reproduction device are received and recorded. The interface 32 is coupled to a central processor unit 33, which in turn is linked to a memory 34. The memory 34 comprises a program memory area 35, in which a program code is stored, via which the operation and operation of the central processor unit 33 is controlled. The program memory area 35 may include both an operating system and an application software that controls the actual image data processing. The memory 34 further comprises a first memory area 36, in which the image data received via the interface 32 are stored. For example, if the image data is read in the form of a 24-bit RGB signal, 24-bit memory is reserved for each pixel. This means that 8 bits are reserved for each of the three basic colors red, green and blue. In addition, it is advantageous to reserve an additional number of bits, for example 8 bits, in the first memory area for each pixel in which an assigned control value can later be stored.

The program code executed on the central processor unit 33 is designed such that control values can be calculated on the basis of two color values, for example based on the color values for the primary colors red and blue, on the basis of a control value function. In a preferred embodiment, control values are calculated for all possible color value combinations and stored in a control value memory area 37 of the memory 34. The control value function may additionally depend on parameters which are dependent on the respective reproduction device for which several partial views of the original view or a spatial image template are generated. In addition, as a parameter Optimum viewing distance from the playback device as a parameter in the control value function. However, since these parameters are constant for all views of an image sequence, for example a video or a television signal, it is advantageous to calculate the control values for all possible combinations once and store them in the control value memory area 37 of the memory 34, instead of calculating the control values individually each time, if these individual pixels are to be assigned, as already explained above.

The program code is further designed such that in each case at least two color values, preferably the color values of the primary color red and the primary color blue, are determined for the individual pixels of the original view, which are stored in the first memory area 36. In the described embodiment, in which the pixels are transferred in an RGB coding, only the 8-bit memory areas which are assigned to the two primary colors have to be read out in each case. On the basis of the values stored therein, which represent the color values for the two primary colors, the corresponding associated control value can now be read from the control value memory area 37 and assigned to the corresponding pixel. This value can then be stored in the reserved bits of the first memory area 36, for example. Alternatively, it is of course possible to store a pointer instead of the value that points to the control value memory area 37. In yet alternative embodiments, it is possible to design the control value memory area 37 in such a way that the color values already present in the RGB coding, which are included in the calculation of the control value, are themselves used as pointer values in a matrix-type memory organization of the control value memory area 37. In a subsequent step, the predetermined shift vectors for the plurality of partial views to be generated, i. the views to be derived, which represent the original scene shown in the original view from different viewing directions generated.

The image lines of the original view are now processed line by line once for each partial view. Depending on the sign of the shift vector, the lines of pixels (here it is assumed that the pixels are organized in horizontal rows and vertical columns) are processed either from left to right or from right to left. The execution direction is determined in each case by the direction of the associated displacement vector.

As already explained above, in each case a disparity vector is formed from the shift vector and the control value associated with the respective pixel, and the Pixels shifted relative to the original view in the generated partial view in accordance with the disparity vector. Here, in some embodiments, it may be provided that the absolute value of the disparity vector is rounded to an integer value, so that a shift takes place about entire grid positions in the matrix-type grid of picture elements. One pixel is shifted by whole column units. In other embodiments, the shift to non-integer positions within the column is permitted and performed, and then an interpolation of the color value is performed to determine the color values at the individual raster points of the image line.

As already explained above, only those pixels in the partial view to be generated are shifted and taken into consideration, which in the execution direction are shifted behind a position (i.e., farther) than a previously shifted pixel. Otherwise, the corresponding pixel is not considered in the partial view. The occurring defects, which are assigned no pixel in the shift, are either filled with the color values, which has a previously correspondingly less far or possibly zero shifted pixel or determined by an interpolation of the color values of the pixels adjacent (not necessarily but immediately adjacent) to the defects within the row. The interpolation may be performed linearly or performed following another predetermined function. In addition, color values of shifted pixels, which are neighbors next to, or next to, etc., can also be taken into account. Since the individual lines of a partial view and the same lines of different partial views can be calculated independently of each other, it is possible to partially or completely parallelize the image processing. As a result, an acceleration of the image data processing can be achieved.

The individual generated partial views of the original scene are stored in second memory areas 38 of the memory 34. If an odd number of partial views are calculated, one of the partial views is usually identical to the original view. Accordingly, in some embodiments, the source view is once copied to a second storage area 38. In some embodiments, the partial views stored in the second storage areas 38 are output via the interface 32 or another interface 39 for further processing and / or storage. In a preferred embodiment, however, the partial views stored in the second memory areas 38 are combined to form a spatial image template, which is temporarily stored in a spatial image memory 40. This summary, which depends on the Playback device to be used can be executed programmatically by means of the central processor device. However, embodiments in which a graphics processor 41 calculates the spatial image template on the basis of the partial views stored in the second memory areas 38 and the output of the spatial image template via the further interface 39 are preferred.

The following is an example of how to create a room image template from the partial views. On the part of the manufacturer of the display device, via which the spatial image template is to be reproduced, the necessary information is generally provided in order to be able to carry out the summary. In the illustrated example it is assumed that the reproduction device 24 different image information in each case different

Viewing direction directions can spend. For the three-dimensional reproduction in the case described here, only four different partial images are used as templates. Since information is to be output in all viewing directions, the same image information is therefore output in several of the spatial directions.

For this exemplary case, the manufacturer of the reproduction apparatus has specified three assignment matrices, which respectively indicate one of the three primary colors red, green, and blue of an RGB color system, such as the corresponding color value of the corresponding base color for the individual pixels from the color values of the pixels of the partial views results. The three assignment matrices for the primary colors green, blue and red are shown correspondingly in FIGS. 6a to 6c. The assignment matrices are each 8x12 matrices. The individual matrix entries each indicate an index associated with one of the four partial views m. The partial view m = 1 is thus assigned, for example in the matrix, the index value 1, etc.

For further explanation, it is assumed that the pixels in the partial views and the spatial image template to be generated are each indexed via an index pair (x, y), where x indicates the position within the line and y the line itself. A pixel of the spatial image template is represented by the expression:

Pixel ^RBV [x, y]

represents. The color value of the basic color "basic color" of the pixel is determined by the expression

Pixel ^RBV [x, y] .primary color

represents, where in an RGB color system "basic color" can assume the values "red", "green" or "blue". The values of the assignment matrices are indexed via an index pair (i, j), where i indicates the position within the line and j the line of the assignment matrix itself. The three color values of a pixel of the spatial image template indexed by the index pair (x, y) result in a pseudo-computer-code notation according to the following instructions:

Pixel ^RBV [x, y] .red: = pixel ^{TA (m)} [x, y] .red with m = assignment matrixRot [i, j],

where i = x modulo 8 and j = y modulo 12,

or shortened:

Pixel ^RBV [x, y] .red: = pixel ^{TA (assignment} matrix ROT ^{[x modul} ° ⁸ ' ^{y modul} ° ¹²¹ > [x, y] .rot, where pixel ^{TA (m)} [x, y] .grunfrabe the color value for Basic color "basic color" of the pixel of the partial view indexed by the index pair (x, y) indicates the partial view m for the other primary colors in the shortened notation corresponding to:

Pixel ^RBV [x, y] .green: = pixel ^{TA (Assignment} matrix GREEN ^{[x modul} ° ⁸ ' ^{y modul} ° ^12] > [x, y]. Green pixel ^RBV [x, y] .blue: = pixel ^{TA Assignment} mat BLUE ^{[x modul} ° ⁸ ' ^{y modul} ° ^12] > [x, y] .blue

The notation "a modulo b" (often also "a mod b"), is defined here by the following calculation rule: where a is a real number and b is a natural number greater than zero and | _cj for the

Gaussian bracket, which gives the real number c the largest integer less than or equal to the number c, i.

| _C_ | : = max (k).

k e Z, k <c

Since the values used for the indexing for x and y in the formulas given above are each integers greater than or equal to zero, the calculation rule: a modulo b: = a - b (a div b) could just as well be used, where (a div b) indicates the quotient a / b rounded to zero.

Using the formulas given above means, for example, that the color values of the pixel on the upper left, for x = 0 and y = 0, are set as follows: The value of the assignment matrix RED [0,0] = 1 can be deduced from the assignment matrix RED. This means that the red color value pixel ^RBV [0,0]. red of the pixel ^RBV [0,0] of the spatial image ^template is given by the red color value pixel ^TA1 [0,0] .rot of the pixel pixel ^TA1 [0,0], the partial view m = 1, the index pair (x, y ) = (0,0). The green color value is also determined by the green color value of the same pixel from the partial view m = 1. (Note m [x, y] = allocation matrix GREEN COLOR [i, j] with i = x modulo 8 and j = y modulo 12)). For the color value of the base color blue, the blue color value of the pixel is used to define the index pair (0,0) of the partial view m = 2 (m = assignment matrix BLUE [0,0] = 2).

Since the assignment matrices are smaller than the fields or the space image template, the matrix entries are used periodically. An assignment matrix that has the size of a partial image or the spatial image template can thus be achieved by juxtaposing and arranging the respective allocation matrix. Mathematically, this can be described by the specified modulo functions. For indices x> 7 and / or y> 11, it is achieved by the corresponding rules i = x modulo 8 and j = y modulo 12 that the corresponding field index of this virtually magnified matrix can be read directly from the given assignment matrix.

4, a further embodiment of a device 51 for image data processing is shown schematically. In this embodiment, the individual functional units are partially or completely formed in hardware, for example by means of an FPGA or an application-related processor chip (ASIC). Via an interface 52, the image data of the original view and optionally further parameters are received again and stored in the first memory area 53. Subviews are subsequently generated by means of a view determination unit 54. Based on a calculation unit 55, the control values are calculated and stored in the control value storage area 56. A color value determination unit 57 determines, for each of the pixels, two color values of two primary colors of a color system. For this purpose, it may be necessary to convert the color information of a pixel from one color system to another color system. On the basis of the determined color values, an assignment unit 58 assigns a control value of the control value memory area 56 to the respective pixel. The assigned control value can also be stored, for example, in the first memory area 53. As already explained above, a partial view determination unit 59 then uses the image data stored in the first memory area and the assigned control values to determine a plurality of partial views which are stored in second memory areas 61. This is a shift of the pixels of the original view in a shift unit 60, as this is explained by way of example in connection with FIG. 2. If appropriate, a summary unit 62 summarizes the multiple partial views into a spatial image template which is optionally stored in a spatial image storage 63 and is then output via a further interface 64. In other embodiments, the intermediate storage of the spatial image template and, if appropriate, the creation of the spatial image template may be omitted. In the latter case, the partial views are then output and provided directly via the further interface.

Depending on how the hardware is designed, it may be advantageous to distribute the computation of the spatial image template when processing from one sequence of views to multiple cycles of that sequence. This is shown by way of example in FIG. 5. In a first image clock, the image data is acquired by a detection module 81 and written to a removable memory buffer 82. This corresponds to the first storage area. In the subsequent image clock, the image data is read from the removable memory buffer 82 and processed by a processing unit as indicated above, by assigning control values from a control value memory 84 to the pixels, and then determining and executing the corresponding displacements of the pixels together with the displacement vectors, and so on Partial views are generated, which are stored in further removable storage buffers 85. In a subsequent image clock, the partial views stored in the further removable storage buffers 85 are then read in as a so-called texture buffer 87 of a graphics processing unit 88. The individual texture buffers 87 are associated with punch masks 89. Via an AND operation between the respective punch mask 89 and the texture buffer 87, the color values of individual pixels for the creation of the spatial image template are selected and combined via OR combinations in a so-called blending unit 90 to form a spatial image template, which in a subsequent image clock of an autostereoscopic reproducing device 91 is output.

The use of the various memory or memory areas may be limited particularly in embodiments implemented in hardware. For example, the calculated partial views or their pixels can be output without buffering. Even if a room image template is to be calculated, it is not necessary to buffer all partial views. It is sufficient to cache only parts, for example parts of the same line of the partial views, in order to determine the individual pixels of the spatial image template. In optimized circuits, a memory requirement can be greatly reduced. 7 schematically shows an embodiment in which the control value assignment to the pixels and the "actual" determination of the partial views take place spatially separated A 2D image data stream 102 is provided to a transmitting station 101. The 2D image data stream 102 comprises, for example, successive original views of one The transmitting station 101 comprises an image data processing device 103, which assigns disparity information in the form of control values to the individual pixels of the 2D image data, as explained in detail above, the device comprising an allocation unit 104 for this purpose Color evaluation unit, a calculation unit and optionally a control value memory include, which are not shown separately.An interface, which is formed for example as a transmitting antenna 105, the 2D image data together with the image assigned control values to one or more receiving stations 106, of which only one is shown, transmitted.

The receiving station 106 comprises an interface formed, for example, as a receiving antenna 107. The 2D image data, along with the control values associated with the pixels, are passed on to an image processing apparatus 108 for producing image data for display, which produces a three-dimensional impression on a viewer. This device 108 is designed to generate the partial views 7-1 to 7-4 and possibly also a spatial image original 10, as described in detail above. The determination of the control values is of course no longer necessary, since these are already transmitted with the 2D image data. The device 108 thus comprises a view determination unit 109, which determines partial views 7-1 to 7-4. Preferably, a spatial image template 10 is created from the partial views 7-1 to 7-4 in a summary unit 110, which is output via an output interface 111 to an autostereoscopic reproduction apparatus 112 and reproduced therefrom.

It goes without saying for the person skilled in the art that only exemplary embodiments are described here. In particular, the process of merging the individual partial views into a spatial image template is described here only by way of example. This is in its exact configuration depending on the respective stereoscopic playback device to be used for playback. LIST OF REFERENCE NUMBERS

1 procedure

2 2D image data

3 data carrier

4 image sequence

5 receiver / player

6 View Discovery Unit

7-m partial view m (m = 1, m)

8 control values

9 playback device

10 room image template

1 1 recombination step

21, 21-k pixel, pixel k (k = 1, 2, ...)

22 line (the original view)

23, 23-k control value, control value of the pixel k (k = 1, 2,

24 line (the shifted pixels)

25, 25-I shifted pixel, shifted pixel I (I

Si process step i

31 Device for image data processing

32 interface

33 processor unit

34 memory

35 program memory area

36 first storage area

37 Control value memory area

38 second memory areas

39 more interface

40 room image storage

41 graphics processor

51 Apparatus for image data processing

52 interface

53 first storage area

54 View determination unit

55 Calculation unit

56 Control value memory area

57 Color evaluation unit allocation unit

Partial view detection unit

shifter

second memory areas

Summary unit

Space template memory

further interface

acquisition module

Removable storage buffer

Control values memory

Removable storage buffers

texture buffer

Graphics Processing Unit

matte

Blending unit

Playback device

transmitting station

2D image data stream

Device for image data processing

allocation unit

transmitting antenna

receiving station

receiving antenna

Image processing apparatus for generating image data which gives a three-dimensional impression upon reproduction

View detection unit

Summary unit

Output interface

Playback device

arrow

Claims

claims

A method (1) of generating image data for reproduction comprising a

creates a three-dimensional visual impression on a human observer, comprising the steps:

Receiving and / or reading in 2D image data of an original view (2), deriving from several further partial views (7-1 to 7-4), the information shown in the original view (2) of different

Represent viewing angles;

characterized in that

for each pixel of the 2D image data of the original view (2) color values of at least two primary colors of a color system are determined, each of the pixels (21-k) is automatically assigned a control value dependent on the at least two determined color values, each of the control values representing disparity information,

and in deriving the plurality of partial views (7-1 to 7-4) representing different viewing directions of an original scene represented by the 2D image data of the original view (2), the individual pixels (21-k) of the original view (2) are pixel-wise corresponding to one for the respective view predetermined displacement vector and the individual pixel associated control value (8, 23-k) relative to each other are moved and based on the shifted pixels (25-I), the pixels of the respective partial view (7-1 to 7-4) are derived ,

2. Method (1) according to claim 1, characterized in that the control values

(8, 23-k) using a control value function that depends on the at least two specific color values as variables.

3. Method (1) according to one of the preceding claims, characterized

characterized in that the pixels (21-k) of the original view (2) are associated with a matrix having vertical columns and horizontal rows, and the displacement vectors are collinear with the rows, such that the individual pixels (21-k) move relative to each other within their Be moved line.

4. Method (1) according to one of the preceding claims, characterized

characterized in that the control value function additionally of parameters depends on an embodiment of the stereoscopic

Playback device (9) and / or optimal viewing distance of the stereoscopic playback device (9) are fixed.

Method (1) according to one of the preceding claims, characterized

characterized in that the control values (8, 23-k) for all possible

Color value combinations of the at least two color values are calculated and stored in a memory (34) and assigning the control values (8,23- k) to the pixels (21-k) from the memory (34 ).

Method (1) according to one of Claims 1 to 4, characterized in that the control values (8, 23-k) are each individually calculated currently for each of the pixels (21-k) on the basis of the control value function.

Method (1) according to one of the preceding claims, characterized

characterized in that the plurality of determined partial views (7-1 to 7-4) are n views, where n is a natural number, and n> 2 and

Displacement vector V _{m of} the view m with m = 1, n is given by

V _m = a * (-1 + 2 * (m-1) / (n-1)) and a is a constant, and preferably a = 1.

Method (1) according to one of the preceding claims, characterized

characterized in that the control value (8, 23-k) for a pixel (21-k) is calculated with the color values ^ and f ₂ for two primary colors according to the following formula:

const.), where const. stands for a normalization constant and fi ^lim and f ₂ " ^m

Color threshold values of the primary colors are at which the control value function S (f ₁ , f ₂ ) has an extremum.

Method (1) according to one of the preceding claims, characterized

characterized in that the pixels (21-k) each according to a

Disparity vector formed of a product of the shift vector associated with the corresponding partial view (7-1 to 7-4) and the control value (23-k) associated with the pixel (21-k), where lengths of the disparity vectors are integer values - and / or rounded off.

10. Method (1) according to one of the preceding claims, characterized in that the individual pixels (21-k) of the original view (2) when generating one of the several partial views (7-m) in each case according to one for the respective pixel (21- k) the disparity vector for each pixel (21-k) from a product of the pixel (21-k) associated control value (23-k) and the shift vector V _{m of} the corresponding partial view (7-m) is calculated, wherein a pixel of the partial view (7-m) is determined on the basis of the color values of the pixel (21-k) to which the higher control value is assigned, provided two pixels (21-k, 21-j) of the original view (2) be moved to the same position or the same shifted pixel (25-I).

11. The method (1) according to any one of the preceding claims, characterized

characterized in that a matrix-like pixel array of the partial view (7-m) is filled based on the displaced pixels (25-I), the color values of array locations being optionally determined by color values of a displaced pixel (25-I) whose position is along the line the direction of the shift vector is before the position of the pixel of the respective partial view to be defined, and / or the color values are calculated by interpolation from the color values of the shifted pixels (25-I) adjacent to the position of the pixel to be defined.

12. The method (1) according to any one of the preceding claims, characterized

in that a pixel resolution of the original view (2) is converted to a pixel resolution of the multiple partial views (7-1 to 7-4) to be generated, before the pixels (21-k) control values (8, 23-k) are assigned.

13. Apparatus (31, 51) for image data processing for generating image data for reproduction via a reproduction apparatus (9), the reproduction of which via the reproduction apparatus (9) produces a three-dimensional visual impression in a human observer, wherein the apparatus (31, FIG. 51) comprises:

an interface (32, 52) for receiving 2D digital image data of an original view (2);

a view determination unit (54) for deriving a plurality of others derived partial views (7-1 to 7-4) representing the original scene represented in the original view (2) from different viewing angles;

characterized in that

the view determination unit (54)

a color value determination unit (57) which determines color values for at least two primary colors of a color system for each of the pixels (21 -k) of the original view (2),

an allocation unit (58) which assigns a control value to each of the pixels (21-k) as a function of the at least two color values determined,

and a shift unit (60) comprising, for each of the further partial views (7-1 to 7-4), each of the individual pixels (21-k) of the

Origin view (2) depending on the associated control value (8, 23-k) and one for the corresponding partial view (7-m) given

Displacement vector (V _m ) shifts and based on the shifted pixels (25-I) derives the pixels of the respective partial view (7-m).

Device (31, 51) according to claim 13, characterized by a calculation unit (55) for calculating the control values (8, 23-k) on the basis of a control value function which is dependent on the at least two determined color values as variables.

15. Device (31, 51) according to any one of claims 13 or 14, characterized

in that the calculation unit is linked to a control value memory and the control values (8, 23-k) for all color value combinations are calculated and stored in the control value memory so that the allocation unit can read out the control values (8, 23-k) from the control value memory.

16. Device (31, 51) according to any one of claims 13 to 14, characterized

in that the allocation unit (58) is designed such that the control values (8, 23-k) are each calculated individually currently for each of the pixels (21-k) on the basis of the control value function.

17. Device (31, 51) according to any one of claims 13 to 16, characterized

in that the control value function is additionally dependent on parameters that are provided by a configuration of the stereoscopic Playback device and / or optimal viewing distance of the stereoscopic playback device are defined.

Device (31, 51) according to one of claims 13 to 17, characterized

characterized in that the plurality of determined partial views (7-1 to 7-4) are n views, where n is a natural number, and n 2 and the displacement vector V _{m are given} the partial view m (7-m) with m = 1, m is by V _m = a ^* (-1 + 2 ^* (m-1) / (n-1)) and a is a constant, and preferably a = 1 holds.

19. Device (31, 51) according to any one of claims 13 to 18, characterized

characterized in that the control value (8, 23-k) for a pixel (21-k) is calculated with the color values ^ and f ₂ according to the following formula:

const.), if otherwise S (f ₁ , f ₂ ) = (| f ₂ -f ₂ ^lim | / const.), where const stands for a normalization constant and where i ^ ' ^m and f ₂ ^{lim are} color limit values of the primary colors in which the control value function S (fi, f ₂ ) has an extremum.

20. Device (31, 51) according to any one of claims 13 to 19, characterized

characterized in that the shift unit (60) is arranged to shift the individual pixels (21-k) according to disparity vectors, each consisting of a product of a product of the shift vector associated with the corresponding partial view (7-1 to 7-4) and the Pixel (21-k) associated control value (23-k) are formed, wherein the shift unit (60) is formed from the lengths of the disparity vectors on integer values up and / or round off.

Device (31, 51) according to one of claims 13 to 20, characterized

characterized in that, when one of the plurality of partial views (7-m) is generated, the individual pixels (21 -k) of the original view (2) are each shifted in accordance with a disparity vector determined for the respective pixel (21-k), the disparity vector for each pixel (21 -k) is calculated from a product of the control value (23-k) assigned to the pixel (21-k) and the displacement vector (V _m ) of the corresponding partial view (7-m), the pixels (21-k) the original view (2) are processed line by line, one by one, in the direction in which the respective displacement vector associated with the partial view (7-m) being generated and pixels (21-k) shifted to a position, which lags behind a position or one Position is similar to that previously a pixel (21-k) of the line (22) has been moved, disregarded.

22. Device (31, 51) according to any one of claims 13 to 21, characterized

in that, when deriving the pixels of the further partial views (7- 1 to 7-4), in each case the pixels (25-I) shifted for the corresponding partial view are taken into account.

23. Device (31, 51) according to any one of claims 13 to 22, characterized

characterized in that a pixel of the partial view m (7-m) is determined on the basis of the color values of the pixel (21-k) to which the higher control value is assigned, provided that two pixels (21-k), 21-1) of the original view (2 ) are shifted to the same position or the same shifted pixel (25-j).

24. Device (31, 51) according to any one of claims 13 to 23, characterized

characterized in that a matrix-like pixel array of the partial view m (7-m) is filled on the basis of the displaced pixels (25-I), the color values of array locations being optionally determined by color values of adjacent displaced pixels (25-I) and / or the color values an interpolation is calculated from the color values of the adjacent displaced pixels (25-I) of the array location.

25. Device (31, 51) according to any one of claims 13 to 24, characterized

in that a pixel resolution of the original view (2) is converted to a pixel resolution of the plurality of partial views (7-1 to 7-4) to be generated, before the pixels (21-k) are assigned control values.

26. Device (31, 51) according to any one of claims 13 to 25, characterized

characterized in that second memory areas are provided in which the partial views are or are stored and a summary unit (61) is coupled to the second memory areas (38, 61) and for creating a spatial image template from the several partial views (7-1 to 7-4 ) is trained.

27. A method for generating image data for a reproduction comprising a

three-dimensional visual impression in a human viewer causing the steps:

Receiving and / or reading in 2D image data of an original view (2) and control values (23-k) associated with pixels (21-k) of the 2D image data, each of the control values (23-k) representing disparity information and for the associated one Pixel (21 -k) of the 2D image data of the original view (2) from color values of at least two primary colors of a color system is determined or can be determined automatically from these, deriving a plurality of further partial views (7-1 to 7-4), the in the Represent original information shown from different viewing angles;

wherein in deriving the plurality of partial views (7-1 to 7-4), the

represent different viewing directions of an original scene represented by the 2D image data of the original view (2), the individual pixels (21-k) of the original view (2) pixel by pixel according to a displacement vector given for the respective view (7-1 to 7-4) the individual pixel associated control value relative to each other are shifted and based on the shifted pixels (25-I), the pixels of the respective partial view (7-1 to 7-4) are derived.

28. Method according to claim 27, characterized in that the pixels (21- k) of the original view (2) are assigned to a matrix having vertical columns and horizontal rows, and the shift vectors are collinear with the rows, so that the individual pixels (21- k) are moved relative to each other within their line when moving.

29. The method according to claim 27 or 28, characterized in that the

n are several views (7-1 to 7-4) n views, where n is a natural number, and n -Ϊ 2 and the shift vector V _{m of} the view m with m = 1 n is given by

V _m = a ^* (-1 + 2 ^* (m-1) / (n-1)) and a is a constant, and preferably a = 1.

30. The method according to any one of claims 27 to 29, characterized in that the pixels (21 -k) are each shifted according to a disparity vector, the product of the corresponding partial view (7-1 to 7-4) associated displacement vector and the The control value (23-k) assigned to the pixel (21-k) is formed, with lengths of the disparity vectors being rounded up and / or down to integer values.

31. The method according to any one of claims 27 to 30, characterized in that the individual pixels (21 -k) of the original view (2) when generating one of the plurality of partial views (7-m) respectively in accordance with a for each pixel (21 -k ) certain disparity vector are shifted, the

Disparity vector for each pixel (21-k) from a product of the pixel (21-k) associated control value (23-k) and the shift vector V _{m of} the corresponding partial view (7-m) is calculated, wherein a pixel of the partial view ( 7-m) is determined on the basis of the color values of the pixel (21-k) to which the higher control value is assigned, provided two pixels (21-k, 21-j) of the original view (2) have the same position or the same shifted pixel (25 -I) are moved.

32. The method according to any one of claims 27 to 31, characterized in that based on the shifted pixels (25-I), a matrix-like pixel array of the partial view (7-m) is filled, wherein the color values of Arraystellen

optionally determined by color values of a shifted pixel (25-I) whose position in the line along the direction of the displacement vector is before the position of the pixel of the respective partial view to be defined, and / or the color values by interpolation from the color values to the position of the pixel to be determined adjacent shifted pixels (25-I) are calculated.

33. Apparatus (108) for image data processing for generating image data for a

Reproduction via a playback device (112), the rendering of which via the playback device causes a three-dimensional visual impression in a human observer, comprising: an interface for receiving 2D digital image data of an original view and control values, wherein each pixel is assigned exactly one control value and each of the control values represents disparity information and is determined for the associated pixel of the 2D image data of the original view from color values of at least two primary colors of a color system of this associated pixel or can be determined automatically from these color values; a view determination unit for deriving a plurality of partial views representing the original scene represented in the original view from different viewing angles; and a summary unit (110) for creating a spatial image template from the plurality of partial views, wherein the viewer determination unit (109) comprises a shift unit that is different for each one the sub-views of each of the individual pixels of the original view depending on the associated control value and for the corresponding partial view (7-m) predetermined displacement vector shifts and based on the shifted pixels derives the pixels of the respective partial view.

34. Device (108) according to claim 33, characterized in that the

a plurality of determined partial views (7-m) are n views, where n is a natural number, and n Ξ- 2 and the shift vector V _{m of} the partial view m (7-m) with m = 1 m is given by

V _m = a * (-1 + 2 * (m-1) / (n-1)) and a is a constant, and preferably a = 1.

35. Device (108) according to claim 33 or 34, characterized in that the

Shift unit (60) is adapted to move the individual pixels (21 -k) according to disparity vectors, each of a product of a product of the corresponding partial view (7-m) associated

Moving vector and the pixel (21-k) associated control value (23-k) are formed, wherein the shift unit (60) is formed from the lengths of the disparity vectors on integer values and / or rounded.

36. Device (108) according to any one of claims 33 to 35, characterized in that the individual pixels (21 -k) of the original view (2) when generating one of the several partial views (7-m) in each case according to one for the respective pixel ( 21), the disparity vector for each pixel (21-k) being made of a product of the control value (23-k) associated with the pixel (21-k) and the shifting vector (V _m ) of the corresponding partial view (FIG. 7-m), wherein the pixels (21-k) of the original view (2) are processed line by line one by one in the direction in which the respective displacement vector associated with the partial view (7-m) being generated is assigned; and disregarding pixels (21-k) that are moved to a position that remains behind a position or equals a position to which a pixel (21-k) of the line (22) has previously been moved.

37. Device (108) according to any one of claims 33 to 36, characterized in that when deriving the pixels of the further partial views (7-1 to 7-4) in each case the shifted for the corresponding partial view pixels (25-I) are taken into account. Device (108) according to one of Claims 33 to 37, characterized in that a pixel of the partial view m (7-m) is determined on the basis of the color values of the pixel (21-k) to which the higher control value is assigned if two pixels ( 21 -k, 21-1) of the original view (2) are shifted to the same position or the same shifted pixel (25-j).

Device (108) according to any one of claims 33 to 38, characterized in that based on the shifted pixels (25-I) a matrix-like

Pixel array of the partial view m (7-m) is filled, the color values of Arraystellen optionally by color values of adjacent shifted pixels (25-I) are set or / and the color values by a

Interpolation from the color values of the adjacent shifted pixels (25-I) of the Arraystelle be calculated.

Device (108) according to any one of claims 33 to 39, characterized in that a pixel resolution of the original view (2) on a

Pixel resolution of the multiple partial views (7-m) to be generated before the partial views (7-m) are generated.

Device (108) according to any one of claims 33 to 40, characterized in that second memory areas are provided in which the partial views (7-m) are or are stored and the summary unit (110) is coupled to the second memory areas and for creating one

Room image template (10) from the multiple partial views (7-m to 7-m) is formed.