EP2540089A2

EP2540089A2 - Method for visualizing three-dimensional images on a 3d display device and 3d display device

Info

Publication number: EP2540089A2
Application number: EP11720977A
Authority: EP
Inventors: Ivo-Henning Naske; Sigrid Kamins-Naske; Valerie Antonia Naske
Original assignee: Expert Treuhand GmbH
Current assignee: Expert Treuhand GmbH
Priority date: 2010-02-25
Filing date: 2011-02-25
Publication date: 2013-01-02
Also published as: US20130147804A1; JP2013520890A; US9324181B2; JP2013527932A; US20130135720A1; JP2017078859A; DE102010009291A1; WO2011103865A2; US9396579B2; JP6142985B2; JP6060329B2; WO2011103866A3; KR20130008555A; KR20130036198A; US20160205393A1; US20160300517A1; WO2011103867A1; WO2011103865A3; WO2011103866A2; US10229528B2

Abstract

A method for visualizing three-dimensional images on a 3D display device, wherein an image to be visualized is supplied as an input image, is characterized in that at least one feature matrix is determined using the input image, the feature matrices defining light/dark information, and in that a display image for reproduction on the 3D display device is produced from the input image using said light/dark information. The invention further relates to a corresponding 3D display device.

Description

METHOD FOR VISUALIZING THREE-DIMENSIONAL IMAGES ON A 3D DISPLAY DEVICE AND

3D DISPLAY DEVICE

The invention relates to a method for visualizing three-dimensional images on a 3D display device, wherein an image to be visualized is supplied as an input image. Furthermore, the invention relates to an SD display device, in particular a stereoscopic or autostereoscopic display, for the visualization of three-dimensional images.

Methods and 3D display devices of the type in question have been known from practice for years. There are stereoscopic devices equipped with vision aids such as red / blue glasses, shutter or polarization glasses etc. allow viewing of a three-dimensional image.

Furthermore, there are autostereoscopic visualization systems that allow one or more viewers, who are in front of an autostereoscopic display, to view a three-dimensional image without visual aids. For this purpose, e.g. Parallax barrier systems or lenticular lens systems positioned in front of the display panel. Since one or more observers may be at different angles relative to a direction perpendicular to the display, more than two perspectives must always be generated and fed to the respective left and right eyes in order to provide as natural a three-dimensional image impression as possible for the viewer allow respective viewer. These systems are also referred to as multi-viewer systems or multiview systems.

In the known 3D display devices, however, is problematic that the quality of the image reproduction for the viewer is not satisfactory.

The present invention is therefore based on the object, a method for the visualization of three-dimensional images on a 3D display device of the type mentioned in such a way and further, that the

CONFIRMATION COPY Visualization of three-dimensional images is improved with simple constructive means. Furthermore, a corresponding 3D display device is to be specified.

According to the invention the above object is solved by the features of claim 1. Thereafter, the present method for visualizing three-dimensional images on a 3D display device is characterized in that using the input image at least one feature matrix is determined, wherein the feature matrices define light / dark information, and that from the input image using the Hell / Dark information, a display image for reproduction is generated on the SD display device.

Furthermore, the above object is solved by the features of claim 17. Accordingly, the SD display device in question, in particular a stereoscopic or autostereoscopic display, for the visualization of three-dimensional images, is characterized in that the 3D display device has means which determine at least one feature matrix using an input image supplied, wherein the feature matrices Define bright / dark information, and that the means from the input image using the light / dark information to generate a display image for playback on the 3D display device.

In accordance with the invention, it has first been recognized that it is of considerable advantage for the visualization of three-dimensional images when one considers that the human eye possesses many times more light / dark receptors than color receptors. In a further inventive manner has been recognized that it is particularly advantageous if one adapts the method or the 3D display device to the anatomy of the eye and the downstream information display. Specifically, at least one feature matrix is determined using the input image, with the feature matrix or feature matrices defining light / dark information. Using the light / dark information is extracted from the Input image generates a display image for playback on the 3D display device. Consequently, the visualization of three-dimensional images is improved with simple constructive means.

Advantageously, the input image may comprise two perspectives - partial images - which correspond to a left and a right partial image.

For ease of processing / editing, the sub-images may be rectified, i. be brought to stereo normal form.

Advantageously, to generate a feature matrix by means of feature extraction, features can be extracted from the input image (s).

The features can describe local properties. For example, shapes, textures and / or edges can be considered as local properties.

In an advantageous embodiment, a feature extraction Sobel feature extraction can be used.

In a particularly advantageous manner, a feature matrix comprises edge information. The human brain essentially uses edges on objects to build up the three-dimensional space image in the brain. Consequently, the edge information greatly facilitates the work of the observer's brain and improves the adaptation to the anatomy of the eye and the downstream information processing in the brain.

In a further advantageous embodiment, feature extraction can be carried out using a Speed up Robust Features (SURF) feature extraction. For further details regarding the SURF process, see H. Bay, T. Tuytelaars and LV Gool, "SURF: Speeded Up Robust Features", Computer Vision and Image Understanding, 110 (3), 2008, pp. 346-359. Advantageously, a feature matrix may include information about prominent pixels.

In a further advantageous manner, a characteristic value can be assigned to a pixel of a perspective of the input image as light / dark information by a feature matrix. Consequently, feature values can be assigned to each pixel of a perspective of the input image as light / dark information.

In an advantageous embodiment, the feature value can be added to the subpixels of the associated pixel in the display image and / or multiplied.

Advantageously, the feature value can be weighted with a scaling factor. In addition, the scaling factor can be changed interactively by a control unit, preferably with a remote control.

In a display device with RGB subpixels, the features of an edge operator used to emphasize the edges in the RGB subpixels can be used. A pixel, e.g. consisting of the RGB subpixels, can be adapted to the anatomy of the eye as follows:

R ^new (i, j): = R (i, j) + M (i, j, 1) - s,

G ^new ("J): = G (i, j) + M (i, j, 1). s

B ^new (U): = B (i, j) + M (ij, 1) - s, where R (i, j), G (i, j) and B (i, j) are the respective colors red, green and define blue. M (i, j, 1) is the value of the edge operator or edge information feature matrix for the pixel (i, j). s is a freely adjustable scaling parameter. When controlled via a remote control, each viewer can adjust the edge enhancement for their own feelings. This method slightly emphasizes the color values of the stereo image in the edges (but not color distorted) and makes them more easily recognizable for the light / dark receptors.

Several features can be integrated as follows:

R ^new (i, j): = R (iJ) + s * ZM, (i, j) * s "

G ^new (i, j): = G (ij) + s * ZM, (i, j) * s "

B ^new (i, j): = B (i, j) + s · M, (i, j) · s "with the feature vectors M (i, j): = (M (i, j, 1) ,. .., M (i, j, K)), where K is the number of different extracted features, and the weighting vector

S: = (s ,, ..., s _K ) for each pixel (i, j). Of course, such an optical enhancement can also be multiplicative.

Furthermore, in a particularly advantageous embodiment, the light / dark information can be displayed with additionally supplemented light / dark subpixels in the display image. The bright / dark subpixels can significantly improve the viewer's spatial image impression by displaying edge information.

For example, as a 3D display device, an autostereoscopic display could be characterized by a panel of subpixels and an upstream optical element. The subpixels can be both color subpixels such as RGB or CMY and bright / dark subpixels. In the color subpixels, the color information of the subpixels of the perspectives to be displayed can be displayed. The light / dark subpixels can contain, for example, grayscale image features that support the impression of the image. Thus, it is considered that the human eye has about 1 10 million light / dark receptors and only about 6.5 million color receptors. In addition, since the human brain uses a substantial amount of edges on objects to construct the three-dimensional space image in the brain, edge information can be displayed via light / dark subpixels, and thus, this image information becomes taken over the much larger number of light / dark receptors. The work of the brain is relieved.

In addition, to improve the quality of the display image, the number of displayed subpixels can be significantly increased. A pseudoholographic display may e.g. contain at least 10 to 20 times as many subpixels as are present in an input image supplied as a stereo image. This greater number of subpixels makes it possible to represent a larger number of pixels per perspective out of the many perspectives that are synthesized. High-definition images and videos of today's generation have about 1920x1080 pixels with 5760 subpixels per line. At a tenfold, and taking into account those subpixels indicating feature information, a display may advantageously have at least 76,800x1080 subpixels. It is taken into account that in the autostereoscopic displays, the assignment of perspectives takes place at the subpixel level. A summary of pixels is not relevant there.

In a specific advantageous embodiment, a stereoscopic display or an autostereoscopic display can be used as the 3D display device.

In a further advantageous embodiment, the input image may have a first perspective and a second perspective, wherein the second perspective is generated by shifting the first perspective by an amount m> 0. Thus, in a first step, the supplied 2D image can be used as a left partial image. However, as a right-hand sub-image, the same 2D image can be used to shift to the right by an amount m> 0. In addition, a disparity matrix of the following kind can be generated:

D _2D ': = {mli = 1 NZ; j = 1 NS}, where NZ is the number of rows and NS is the number of columns. As a result, all viewers have the impression that the 2D image (which is still perceived as such) floats in front of the display at a certain distance.

If the 2D image were shifted to the left by a certain amount m <0, then the viewers would have the impression that the image is shifted inwards into the display.

Furthermore, one could let the viewer choose the amount m interactively, e.g. By means of a remote control, the viewer can adjust the "pop-out" or "pop-in" effect at any time.

Advantageously, the 3D display device may comprise subpixels comprising subpixels for representing a color of a predeterminable color system and light / dark subpixels for displaying feature information.

Furthermore, the subpixels can be advantageously designed as independent elements.

Advantageously, the subpixels can have the same extent in the horizontal and vertical directions. In a particularly advantageous manner, the subpixels are square, whereby a higher resolution can be achieved. Also conceivable is a round design of the subpixels. Thus, the hitherto customary requirement that all subpixels of a pixel together form a square is dropped. Rather, each subpixel is an independent element. Each of these subpixels has a color of the selected color system and has the same extent in the horizontal and vertical directions. With the OLED or Nano technology, this is technically easy to implement.

There are now various possibilities for designing and developing the teaching of the present invention in an advantageous manner. For this purpose, on the one hand to the claims 1 and 17 subordinate claims and on the other hand to refer to the following explanation of a preferred embodiment of the invention with reference to the drawing. In conjunction with the explanation of the preferred embodiment of the invention with reference to the drawing Also generally preferred embodiments and developments of the teaching are explained. In the drawing show

1 shows a block diagram for illustrating the overall system according to an embodiment of the method according to the invention and the 3D display device according to the invention,

2 is a flowchart of the overall system of FIG. 1,

Fig. 3 is a conventional subpixel layout compared to a new one

Subpixel layout according to an embodiment of a 3D display device according to the invention,

4 shows a conventional subpixel layout compared to a new one

Subpixel layout according to another embodiment of a 3D display device according to the invention and

FIG. 5 shows the subpixel layout of FIG. 4, wherein a larger number of different perspectives are activated.

1 shows a block diagram for illustrating the overall system according to an exemplary embodiment of the method according to the invention and the 3D display device according to the invention. The exemplary embodiment according to FIG. 1 relates to a method for the visualization of three-dimensional images on an autostereoscopic display as a 3D display device, on which a plurality of perspectives, generally more than 100, from a supplied stereo image in any 3D format are interlaced on the display are displayed. The display consists of an optical element and an image-forming unit. The multitude of perspectives is generated in such a way that only those pixels of a perspective are generated, which also have to be displayed. The imaging unit of the display consists of subpixels emitting a color, eg red, green or blue. The autostereoscopic display is capable of receiving a 3D image or an SD image sequence in any format, such as a stereo image or a stereo image sequence. Other formats, such as stereo image including a disparity card, can also be received and processed.

A received stereo image is first rectified, i. brought to stereo normal form or in the epipolar standard configuration. If this is already the case, then the identical figure arises here.

Subsequently, according to FIG. 1, various features such as edges or prominent pixels (SURF) are identified in the two images R 1 and R _r . Various features can be extracted here. A special restriction is not given. These features are used both for calculating the disparity map, if required, and for visualization in certain additional subpixels.

If the disparity card has already been received with the input image, the next step is skipped. Otherwise, the disparity map of the stereo image is calculated. It contains an assignment of the pixels of the left and right field, which are present in both received perspectives. In addition, the left and right occlusions are identified.

By means of the disparity map D of the received stereo image and the features then any number of perspectives are synthesized. This is done so that only those subpixels are synthesized, which must be displayed on the display actually. Thus, for 100 perspectives to be displayed, only 1% of the subpixels are calculated from each perspective.

The information about which perspective is to be displayed on which subpixel is defined in the perspective map P. The perspective map is defined and stored in the production of the display by a calibration process between subpixels and optical system. Adjacent subpixels are generally associated with different perspectives. The storage of different subpixels from the different perspectives in the pseudoholographic image B is referred to as mating.

The autostereoscopic display is characterized by a panel of subpixels and an upstream optical element. The subpixels are color subpixels, e.g. RGB or CMY. In the color subpixels, the color information of the subpixels of the perspectives to be displayed is displayed.

In addition, to improve the quality of the image B, the number of displayed subpixels is substantially increased. A pseudoholographic display according to FIG. 1 has at least 10 to 20 times as many subpixels as are present in the received stereo image. This greater number of subpixels makes it possible to represent a larger number of pixels per perspective out of the many perspectives that are synthesized.

FIG. 2 shows a flowchart with the individual steps to the overall system from FIG. 1.

The steps according to FIG. 1 and FIG. 2 are described in more detail below. In the context of the exemplary embodiment, it is assumed that a picture sequence of stereo images is received, decoded and made available in the memory areas I, and I _r by a receiving module, for example via an antenna or the Internet.

After receiving the stereo image is enlarged or reduced to the resolution of the connected pseudo-holographic display. A display with a resolution of 19,200 x 10,800 pixels can be considered high-resolution. In this case, e.g. a stereo HD image enlarged tenfold horizontally and vertically.

In a first step, the rectification is performed. These methods are known from the literature. In the exemplary embodiment according to FIGS. 1 and 2, in the left partial image I, nine prominent points, which are distributed uniformly over the image, are searched for by the SURF method. The coordinate of each distinctive Point is used as the center of a search block in the right field l _r . In this search block, the most similar point in the right field l _{r is} searched. By means of the disparities of the nine distinctive points is defined a linear transformation matrix, with which each of the partial images I, and l _r to R, and R _r rectified:

/, -> R, and \, -> R _r

With this, the rectification is carried out, so that now the epipolars can run parallel to the lines of the image and all subsequent operations can be performed line by line.

For calculating features, e.g. the SURF (Speed up Robust Features) or the Sobel Edge Detector method are used.

SURF feature extraction:

This method is based on approximating the determinant of the Hesse matrix for each pixel. The procedure is as follows.

In a matrix Isum, which is created for the left and right rectified partial image, the partial sums of the gray values of the lower image area are stored for each pixel: süm (/, y ^' ): = ^ X / (/ ^■,; ^' )

/, = o y, = o

with i: = 1, ... NZ and j: = 1, ... NS, where NZ is the number of rows and NS is the number of columns.

In the calculation, the procedure is such that first of all each row i is assigned a row calculation unit i which calculates the partial sum recursively: Isum, (y): = Isum, (y - i) + R _r (/ ^' , y)

for i: = 1 .... NZ and j: = 1, .. NS and stored in Isum. Thereafter, each column j is assigned a computation unit j, which computes the column sum recursively: lsum (i, j): = lsum (j-1, j) + lsum _t (y)

for i: = 1 .... NZ and j: = 1 ... NS.

The matrix thus calculated now contains the desired partial sums for each field left and right. Now each line i is again a computing unit i

(= Row calculation unit i), which first calculates the following intermediate values:

^D xx ('· y) ^: = lsum (i - 4, j - 5) - lsum (i + 2 - 5)

+ lsum (i + 2, j ^' + 4) - lsum (i - 4, j ^' + 4)

- lsum (i + 2, j + 1) + 3 · lsum (i + 2 - 2)

+ 3 · lsum (i - 4, j ^' + 1) - 3 · lsum (i - 4, j - 2)

^D yy (i _> y) ^{: =} isum (i + 4, y + 2) - isum (i + 4, j - 3)

+ lsum (i - 5, j + 2} + lsum (i - 5, j - 3)

- lsum (i + 1, y ^' + 2) + 3 · lsum (i + 1, j - 3)

+ 3 · lsum (i - 2, y + 2) - 3 · lsum (i - 2, j - 3)

D _RO (/ ^' , y): = lsum (i + 3, j + 3) - lsum (i + 3, y)

- Isum (i, y + 3) + lsum (i, j)

D _LO (/ ^' , y): = isum (i + 3, j - 1) - isum (i + 3, y - 4)

- Isum (i, lsum (i, j - 4)

D _LU (, y): = lsum (i - 1, y - 1) - lsum (i - 1, y - 4)

- / st m (/ ^' - 4, y - 1) + / st m (/ ^' - 4, y - 4) ^D RU ('> J) ^' ■ = ^lsum (i - 1, y + 3) - lsum (i - 1, j)

- Isum (i - 4, y + 3) + lsum (i - 4, j)

After that arises

£> y (/ ^' . Y): = o _to (, j) + (/, j) ^■ D _RO (/, y) - D _LU (, y)

The approximated determinant of the Hessian matrix is then stored for each pixel of the right field in the feature matrix M _r (1):

M _r (/, y ^' , 1): = (/, j) ^■ Dyy (, j) - 0.81 · (/, j) ^■ (, j)

The same procedure is performed for each pixel of the left field R, and stored in the feature matrix M, (1). Overall, the feature matrices M _r (1) and M, (1) result. For further details, see H. Bay, T. Tuytelaars and LV Gool, "SURF: Speeded Up Robust Features", Computer Vision and Image Understanding, 1 10 (3), 2008, pp. 346-359.

Sobel feature extraction:

The Sobel operator is just one of a large number of edge operators and is therefore described as an example.

For the disparity map, an edge operator is of particular importance, as it helps to assign more importance to edges than to smooth surfaces. Since an edge is always a regional property, this procedure within a row also allows to take into account the properties of local regions.

For example, the Sobel-Prewitt operator works with 3x3 matrices that detect the edges in different directions. Basically, here are horizontal, vertical, left and right diagonal edges to distinguish. For their detection, the following 3x3 matrices are used:

An implementation according to an embodiment of the method according to the invention proceeds in such a way that each row i is assigned a row calculation unit i. For all rows i and columns j, the computational-local field edge (1) to edge (9) is filled from the right-rectified partial image R _r as follows:

Edge): = R _r (i-Xj-i) edge {2): = R _r {j- \ j) edge (3): = F? _r (-1, y ^' + l) edge (4): = R _r (i, ji) edge i): = R _r {i, j) edge (6): = R _r (i, j + i) edge (7): = F? _r (+ 1, / - 1) edge fissure): = R _r (i + Xj) edge (9): = R _r (i + j + i) for i: = 1 ..... NZ andj: = 1 ..... NS Then calculate each row calculation unit i for each index j:

Η _Λ : = 2 · edge (2) + 2 · edge (5) + 2 · edge (8)

- Kantet) - edge (4) - edge (7)

- edge (3) - edge (6) - edge (9) H ₂ : = 2 ^■ edge (A) + 2 · edge (5) + 2 · edge {6)

- edge (1) - edge (2) - edge (3)

- edge (7) - edge (8) - edge (9)

H ₃ : = 2 - edge (7) + 2 ^■ edge (5) + 2 ^■ edge (3)

- edge (1) - edge (2) - edge (4)

- edge {8) - edge leg) - edge (6)

H ₄ : = 2 - edge (1) + 2 · edge (5) + 2 ^■ edge {9)

- edge (2) - edge (3) - edge (6)

Edge (4) edge (7) edge (8)

M _r (i, j, 2) then results as M _r (i, j, 2): = Η + H ₂ + H ₃ + H ₄ for i = 1, ..., NZ and j = 1,. ..NS.

This process is carried out in the same way for the left-hand rectified partial image R 1. Overall, the feature matrices M _r (2) and M, (2) result.

3 shows on the left a conventional pixel layout with the three subpixels R (red), G (green) and B (blue). With these subpixels, the three perspectives 1, 2 and 3 are operated using a lenticular lens as optical element O. FIG. 3 shows a new subpixel layout on the right, wherein the autonomous subpixels according to an embodiment of the 3D display device according to the invention as a autostereoscopic display a square Design have. 9 perspectives can be controlled by the optical element O with 9 subpixels.

FIG. 4 once again shows a conventional pixel layout on the left. In Fig. 4 right another embodiment of an SD display device according to the invention is shown as autostereoscopic display. There, a much finer and more detailed subpixel structure is created. Instead of three subpixels in the conventional pixel layout, 144 subpixels are generated in the subpixel layout of the embodiment. The subpixels R (red), G (green) and B (blue) are another subpixel W (eg white or yellow) for Presentation of light / dark information added. With these 144 subpixels, 36 perspectives are activated in the illustrated embodiment.

FIG. 5 shows the subpixel layout of FIG. 4, wherein the 144 individual, independent subpixels are used to drive 144 perspectives.

According to a further embodiment of the method according to the invention and a 3D display device according to the invention, the procedure for an autostereoscopic display can be as follows.

To adapt to the anatomical conditions of the human eye, two properties are considered. These are:

1. the resolution of the eye and

2. the number and characteristics of the receptors in the eye.

The resolution of the human eye is i.A. between 0.5 'and 1 .5'. Therefore today's displays usually have a dot pitch of 0.2 to 0.3 mm. That is, from a distance of about 1 m, the pixels of the display are no longer visible. In the embodiment of a 3D display device according to the invention presented here, the lens width of the lens grid used is in the range of 0.2 mm. That is about 125 LPI (lenses per inch). As a result, the lens structure is no longer recognizable from a viewing distance of about 1 m. The number of subpixels behind a lens is on the order of 10 subpixels per lens. That is, the dot pitch of a pseudo-holographic display is on the order of about 0.06 mm. While in conventional displays, e.g. 1920 x 1080 pixels (HD-TV), a pseudo-holographic display presented here consists of at least 19,200 x 1080 pixels.

The lenticular may consist of lenticular lenses or hexagonal lenses, for example, which in this case have a diameter of 0.2 mm. Considering that in the human eye about 6.5 million color receptors are present and about 100 million light / dark receptors, it can be seen that about 15 times as many light / dark receptors are available as color receptors. Furthermore, the human brain uses the edge information to a great extent for the generation of the inner space image. Edges give information about the front / back relationship of objects with the existing right and left covers.

In one exemplary embodiment, therefore, the known subpixels RGB or YMC are particularly advantageously supplemented by light / dark subpixels which indicate the edge information generated in the feature extraction phase by the edge operators.

Homogeneous surfaces contain no edges. The light / dark subpixels do not represent any information in the image.

Edges are brightened in the light / dark subpixels according to the intensity of the detected edge. As a result, the edges present in the image are highlighted and more easily recognized by the 100 million light / dark receptors.

The brain has it so easier to create the inner space image. Patterns on homogeneous surfaces are recognized as such in the human brain due to the learning effect and do not affect the spatial impression.

The geometric arrangement of the light / dark subpixels can be varied according to the invention. In Fig. 3, Fig. 4 and Fig. 5, various divisions are shown. In addition to the light / dark subpixels for improving the edge detection, further features for spatial image generation can be added. Here is an example, but not exclusive, called the SURF operator.

In addition, the division of pixels into color subpixels (usually RGB subpixels) is canceled. Taking into account that the assignment of perspectives for an autostereoscopic display always at the level of subpixels with color and Light / dark sub-pixels takes place, so in the display described here, the grouping of sub-pixels to pixels is omitted. Each color or light / dark subpixel is in itself a self-contained light element associated with a particular perspective and has the same extent in the horizontal and vertical directions. This is already taken into account in FIGS. 3, 4 and 5 on the right-hand side. A backward compatibility is still given, so that all 2D images and videos can be displayed easily. The production of such an autostereoscopic display is possible based on the OLED technology.

With regard to further advantageous embodiments of the method according to the invention and of the display device according to the invention, reference is made to avoid repetition to the general part of the description and to the appended claims.

Finally, it should be expressly understood that the above-described embodiments of the method according to the invention and the device according to the invention are merely for the purpose of discussion of the claimed teaching, but do not limit these to the exemplary embodiments.

Claims

A method for visualizing three-dimensional images on an SD display device, wherein an image to be visualized is supplied as an input image,

That is, at least one feature matrix is determined using the input image, wherein the feature matrices define bright / dark information, and that a display image is generated from the input image using the light / dark information for display on the 3D display device.

2. The method according to claim 1, characterized in that the input image comprises two perspectives - partial images - which correspond to a left and a right partial image.

3. The method according to claim 2, characterized in that the partial images are rectified.

4. The method according to any one of claims 1 to 3, characterized in that for generating a feature matrix by means of a feature extraction features from the input image and the sub-image are extracted.

5. The method according to claim 4, characterized in that the features describe local properties.

6. The method according to claim 4 or 5, characterized in that a feature extraction Sobel feature extraction is used.

7. The method according to any one of claims 1 to 6, characterized in that a feature matrix comprises edge information.

8. The method according to any one of claims 4 to 7, characterized in that the feature extraction is a Speed up Robust Features (SURF) feature extraction is used.

9. The method according to any one of claims 1 to 8, characterized in that a feature matrix information about prominent pixels.

10. The method according to any one of claims 1 to 9, characterized in that a feature value is assigned to a pixel of a perspective of the input image as light / dark information by a feature matrix.

1 1. A method according to claim 10, characterized in that the feature value is added to the subpixels of the associated pixel in the display image and / or multiplied.

12. The method according to claim 10 or 1 1, characterized in that the feature value is weighted with a scaling factor.

13. The method according to claim 13, characterized in that the scaling factor by a control unit interactively, preferably with a remote control, is changeable.

14. The method according to any one of claims 1 to 13, characterized in that the light / dark information with light / dark subpixels are displayed in the display image.

15. The method according to any one of claims 1 to 14, characterized in that a stereoscopic display or an autostereoscopic display is used as a 3D display device.

16. The method according to any one of claims 1 to 15, characterized in that the input image has a first perspective and a second perspective, wherein the second perspective by shifting the first perspective by an amount m> 0 is generated.

17. 3D display device, in particular a stereoscopic or autostereoscopic display, for the visualization of three-dimensional images, in particular for carrying out a method according to one of claims 1 to 16,

characterized in that the 3D display device comprises means for determining at least one feature matrix using a supplied input image, the feature matrices defining light / dark information, and the means from the input image using the light / dark information for displaying a display image on the 3D display device.

18. 3D display device according to claim 17, characterized in that the 3D display device has subpixels which comprise subpixels for representing a color of a predefinable color system and light / dark subpixels for displaying feature information.

19. 3D display device according to claim 18, characterized in that the subpixels are formed as independent elements.

20. A 3D display device according to claim 18 or 19, characterized in that the subpixels in the horizontal and vertical directions have the same extent, in particular are designed square.