WO2007052191A2

WO2007052191A2 - Filling in depth results

Info

Publication number: WO2007052191A2
Application number: PCT/IB2006/053944
Authority: WO
Inventors: Jeroen Smit; Richard P. Kleihorst
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2005-11-02
Filing date: 2006-10-26
Publication date: 2007-05-10
Also published as: WO2007052191A3

Abstract

The invention relates to depth determination of an imaged area (118), in particular to improving a density of a depth map (216, 312) of the area (118) by changing and filling in (224) changed depth results. When determining the depth of or distance to one or more pixels representing an object in an area, e.g. only boundaries (126, 128, 130, 132) of the object may be accurately determined, providing a relative sparse depth map (116). Therefore, an improved way is disclosed for determining a changed depth level of one or more pixels in order to provide a relative dense defined depth map from a relative sparse defined depth map. A test for whether to change the depth of a current pixel (306, 316) uses depth level criteria and/or texture level criteria. The invention may e.g. be used with stereovision for 3D user interface applications.

Description

Filling in depth results

FIELD OF THE INVENTION

The invention relates to depth determination of an imaged area, and in particular to improving a density of a depth map of the area by changing and filling in the changed depth results.

BACKGROUND OF THE INVENTION

Obtaining depth level information of a real world scene is one of the most vital visual tasks which humans can do almost effortlessly while for e.g. vision systems and computers it is a difficult and challenging task. However, depth determination systems that perform depth maps e.g. in real time are in high demand, especially looking at the latest developments of consumer robotics, mobile telecommunications or 3D user interface applications.

Depth estimation by vision can e.g. be performed by stereovision. Stereovision provides a robust way to obtain depth information. Stereovision relies on one fundamental finding. If two images of a given area or a scene are captured from two different viewpoints, then the resulting images will differ slightly due to the effect of a perspective projection. The correspondences of a stereo pair can be used effectively to obtain or construct the three- dimensions of the area, via a procedure known as stereo matching. In stereo matching e.g. a distance or e.g. a number of columns that coordinates of an object or one or more pixels of an object in one image are shifted with respect to the same object in the other image, relative to its local coordinate system, is used to obtain depth information of the area.

Some available depth determination systems are able to determine depth in real time, but these are e.g. expensive, power consuming and large, which is undesirable for the mentioned applications. Still further, existing depth determination methods, devices and systems may not provide depth maps in an efficient way, where the depth maps are accurate enough and acceptable for the application.

SUMMARY OF THE INVENTION

The present invention seeks to provide an improved way of changing a depth level of one or more pixels. Preferably, the invention alleviates, mitigates or eliminates one or more of the above or other disadvantages singly or in any combination. The invention is defined by the independent claims. The dependent claims define advantageous embodiments.

The invention thus provides a method for changing a depth level of one or more pixels in dependence on the depth level map and/or the texture level map. In one aspect, a depth level of a current pixel is changed to a depth level of a previous pixel when certain criteria as to the texture level and/or the depth level are fulfilled.

The depth level map obtained before changing the depth level of one or more pixels may be referred to as an initial depth map. The initial depth map may be inaccurate, at least in certain parts of the depth map, and therefore the initial depth map may also be called a sparse depth map. The sparse depth map may be obtained by depth matching using stereovision, where the stereovision is provided by the left and right image.

The sparse depth map is normally only accurate at certain image features of the map, for instance on edges of objects within the area, which edges can be accurately matched, but is not accurate inside an object where no edges are present or is not accurate between edges.

A reason for the sparse depth map normally not being accurate between edges may be that the sparse depth map was approximated, e.g. inside objects, by trying to match on the textures of the objects between the edges. However, a good quality match may not be found on texture information. If there is no texture in an object, the actual depth may never be found using a local matching on texture. Therefore the sparse depth map may result in a depth map where only edges of the objects within the area have an accurately defined depth. One possible advantage obtained by the present invention and in particular by changing a depth level of a current pixel to a depth level of a previous pixel in the depth level map when one or more texture level criteria of the current and previous pixels in the texture level map and one or more depth level criteria of the current and previous pixels in the depth level map are fulfilled, and filling in the changed depth level in the depth level map, may be that a dense depth map may be provided. The dense map may be provided by changing the depth values and filling in the changed values, and thereby providing the dense depth map from the sparse depth map, in which dense depth map also the depth of pixels between edges of an object are accurate. In an advantageous embodiment, a dense map is obtained starting from a sparse map by processing the pixels starting from the pixels for which reliable depth information is available, such as for pixels in texture-rich areas such as at edges. When, after the initial depth map is provided, changing the depth of a current pixel to the depth of a previous pixel, when certain criteria of the texture and depth level of the current and previous pixels are fulfilled, one possible advantage may be that at least the depth of pixels between edges may be filled with depth levels that are provided from (or based on) depth levels of pixels that are accurately depth determined from the edges and which filled in depth levels therefore is more accurate than the depth values in the sparse depth map.

One possible advantage by the present invention is that it has been found that certain criteria, such as difference, for both the texture level and the depth level between one or more current pixels and one or more previous pixels, are accurate indicators of whether to change the depth of the current pixel into the depth of the previous pixel, and therefore possibly changing the accurateness of the depth for the current pixel and filling in the changed depth value in the sparse depth map, and hereby providing the dense depth map.

Comparing the texture levels and the depth levels of the current and previous pixels in the texture level map and depth level map with each other and/or with a threshold may provide testing the fulfillment of the criteria.

A determination of whether one or more texture level criteria of the current and the previous pixel in the texture level map and one or more depth level criteria of the current and the previous pixel in the depth level map are fulfilled may normally be tested for each pixel in the depth map, whereas only the depth level of pixels which fulfils the criteria are changed.

Possibly, when only one of the texture level criteria or one of the depth level criteria is not fulfilled there may be no reason to determine whether another criterion is fulfilled. Instead, the determination of whether one or more texture level difference criteria between the current and the previous pixel in the texture level map and one or more depth level difference criteria between the current and the previous pixel in the depth level map are fulfilled may be provided for a next pixel. A possible advantage by not testing two criteria when a first criterion is false may e.g. be to save processing time.

Changing the depth level of the pixels and filling in the changed depth level in the depth level map is normally an iterative process, so the steps 'changing' and 'filling in' will normally be carried out for a number of iterations, e.g. on a pixelrow-by-pixelrow basis. Furthermore changing and filling in is based on the depth levels, which depth levels may have been changed during the iterative process. As a texture level, an intensity level or an average intensity or level range or deviation of the one or more pixels can be used. Obtaining or changing the depth of a pixel may be interpreted as determining a distance to a part of the area being represented by the pixel. When, in an embodiment of the invention, the one or more texture level criteria comprises the difference between the texture level of the current and previous pixels and wherein the one or more depth level criteria between the current and previous pixel comprises the difference between the current and previous pixels, one possible advantage may be that simple but yet effective criteria for testing whether the depth is to be changed are provided.

When, in an embodiment of the invention, the depth level of the current pixel is changed to the depth level of the previous pixel when a difference between the texture level of the current and previous pixel is smaller than a threshold and a difference between the depth level of the current and previous pixels is smaller than a threshold, one possible advantage may be that only the depth of pixels within the same object may be changed. In other words, an advantage may be that objects in front of the background are filled in based on information from the texture and depth level map. Having a texture difference smaller than a certain threshold may be due to the pixels being within the same object.

When, in an embodiment of the invention, the depth level of the current pixel is changed to the depth level of the previous pixel when the depth level of the previous pixel is lower than the depth level of the current pixel, one possible advantage may be that objects in the foreground are completed and depth values which brings the objects nearest to the cameras are chosen because bringing or "pulling" objects towards the cameras in order to have the lowest depth may introduce less artifacts. When, in an embodiment of the invention, the depth level is changed when a difference in the texture level between the current and the previous pixels is smaller than a threshold and the difference in the depth level is undefined, one possible advantage may be that when the depth value of one or more pixels in the sparse depth map is even not defined due to the depth matching not being able to achieve a good match, or the depth being defined to a value that represents an undefined depth, the present invention may change the depth of the pixel and fill in the changed depth of the pixel in the depth map. The depth of such undefined pixel(s) will then be changed to a depth defined on the basis of accurate or well- matched pixel(s) with an accurate or well-defined depth. When, one texture level criterion and one depth level criterion are determined as differences between local areas of the current and previous pixels, one possible advantage may be a relative smooth defined depth of e.g. the object.

When, in an embodiment of the invention, one texture level criterion and one depth level criterion are tested for local areas of the current and previous pixels and wherein the local area of the current and previous pixels intersects for a number of intersection pixels and wherein the depth level of the current pixel is changed to a depth level of one or more of the intersection pixels, may be that another way of smoothening the depth of e.g. the object is provided. The criteria may be tested for differences in the texture and depth level between the local areas of the current and previous pixel. When the depth level is changed into a level of more pixels, it may be understood as e.g. an average of the depth of these pixels.

When, in an embodiment of the invention, the depth level is changed to the lowest depth level, one possible advantage may be that, objects are preferred to be in the foreground having the lowest depth, as this may introduce less artifacts e.g. for robotics and coding.

When, in an embodiment of the invention, the depth level of the current pixel is changed when a difference in the texture level between the local areas of the current and the previous pixels is smaller than a threshold and a difference in the depth level is undefined, one possible advantage may be that when the depth value of one or more pixels in the sparse depth map is even not defined due to the depth matching not being able to achieve a good match, or the depth being defined to a value that represents an undefined depth, the present invention may change the depth of the pixel and fill in the changed depth of the pixel in the depth map. The depth of such undefined pixel(s) will then be changed to a depth defined on the basis of accurate or well-matched pixel(s) with an accurate or well-defined depth. The small difference in texture may guarantee a high possibility of filling in an object.

When, in an embodiment of the invention, the depth level of the current pixel is changed when a difference in the texture level is smaller than a threshold and a difference in depth level is smaller than a threshold, one possible advantage may be that only the depth of pixels within the same object may be changed. In other words, an advantage may be that objects in front of the background are filled in based on information from the texture.

When, in an embodiment of the invention, at least an inside (or middle) of an object within the area is at least partly filled with changed depth levels in a direction from a boundary of the object towards a middle of a surface of the object, one possible advantage may be that the surface of the object, or the one or more pixels representing the surface, will be provided with a depth defined on the basis of accurate or well-matched pixel(s) with an accurate or well-defined depth. The accurate or well-matched pixel(s) with an accurate or well-defined depth is normally pixels representing the boundaries and in particular the vertical boundaries of the object. These pixels can be identified from an edge picture variant of the left and right images or a quality of the depth matching.

The invention may e.g. be used with stereovision for 3D user interface applications, robotics or mobile telephones. In general the various aspects of the invention and the possible advantages hereby may be combined and coupled in any way possible within the scope of the invention. These and other aspects, features and/or advantages of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention will be described, by way of example only, with reference to the drawings, in which

Fig. 1 shows a prior art system for obtaining a depth level map and a depth level map achieved by the prior art system;

Fig. 2 shows a principle figure of a system according to the present invention;

Fig. 3 shows a detail of a level texture map and a depth level map according to the invention.

DESCRIPTION OF PREFERRED EMBODIMENTS

Fig. 1 shows a stereovision system 102 with a left camera (L. C.) 104 and a right camera (R. C.) 106. The left and right cameras are providing a left image 108 and a right image 110 of an area 118. The images 108 and 110 are inputted to a control logic 114, e.g. formed by a suitably programmed microprocessor, for depth matching (D. M.) 112. As an output of the depth matching a depth map 116 is obtained.

For depth determination by depth matching in an example of a case where cameras are provided at the same height (epipolar principle) and vertical angle, a match of one or more pixels in the left and right images is only searched leftwards or rightwards from a line in the image. l(i) is a pixel from line x in the left image and r(i) is the same pixel on the right image. Note that because the cameras are e.g. 6 cm apart, the pixels do not represent a same part of the image. To determine the depth or distance to the same part of the image the right image data is shifted leftwards over a distance s until a good match is found and l(i) equals r(i-s). Normally a search for a good match is not carried out for only one given pixel, but for a neighborhood, for instance 3x3 or 5x5 pixels around l(i) and r(i) (or r(i-s)). S is shifted e.g. from s=0 until s=200 for a VGA sized image. The s value, s*, that gives the best match is chosen. If the value, s*, is small the pixel (or local group or area of pixels) is far away, (the depth is high). If s* is big, the pixel is close to the cameras. A difference between s* and the other s's give a measure of the quality of the match.

After this, s*(i) is provided for every pixel in e.g. this line by a depth estimate or depth determination according to the described being found. The determination is aligned with l(i) and in s*(i) and l(i) the position i points at the same part of the image. Because s*(i) is only accurate at certain image features of the signal such as edges of objects, only a sparse depth map is hereby provided because only e.g. vertical edges of objects can be accurately matched. The depth map 116 may therefore be referred to as the sparse depth map. The sparse depth map 116 may provide a representation of the depth of two objects 120 and 122 as shown on a background 124. The object 122 is surrounded by edges 126, 128, 130, 132. In this example the edges are outer surrounding edges or boundaries of the object, but edges may off course also be present inside the surrounding edges of the object. A thick black line such as the vertical edges 126 and 132 indicates that these edges are accurately matched and the depth of these edges are therefore accurately determined or estimated in the sparse depth map. The horizontal thin line 130 indicates a horizontal edge of which the depth may be e.g. less accurate determined. The dashed horizontal thin line 128 indicates an even less accurately determined horizontal edge. Between the edges of objects, such as between the edges or boundaries 126,

128, 130, 132 or inside the edges as indicated with the reference number 134 pointing to the area inside the object 122, there has not been found a depth, of one or more pixels, with a preferred accuracy. This inside or middle part of the object may therefore be referred to as a relative sparse defined part of the sparse depth map 116. This may be due to the system not having been able to match one or more pixels accurately enough. This may be due to a relative monochrome surface or the non-presence of vertical and/or horizontal edges that are clear enough to provide a good match and thereby a good depth determination.

Possibly the system has not been able to determine the depth of one or more pixels in the area 134 or a similar area e.g. inside the object 120 because the depth level was approximated inside objects by trying to match on the textures of the objects. However, texture looks different through both cameras and a good quality match may not be found. If there is no texture in an object, the actual depth may never be found using local matching.

Fig. 2 shows an embodiment in accordance with the present invention. Fig. 2 shows a stereovision system 202 with a left camera (L. C.) 204 and a right camera (R.C.) 206. The left and right cameras are providing a left image 208 and a right image 210. The images 208 and 210 are inputted to control logic 214 for depth matching (D. M.) 212 and a filling in process of changed depth results (F.) 224. As an output from the depth matching a depth map 216 (the sparse depth map) is obtained and serves as an input to the filling in process 224 of changed depth results. Furthermore, a texture level map 220, such as a map of the intensity of the pixels in the right image 210 and/or of the left image as indicated by the dashed arrow at 222 is also inputted to the filling in process 224. The output of the depth matching and the filling in process 224 is a dense depth map 226. The dense depth map 226 accurately defines the depth of a complete visible surface of e.g. the objects 120 and 122 shown in FIG. 1, and e.g. not only of the edges of the objects.

The dense map is provided by comparing a texture level, such as an intensity level difference, of a current pixel and a previous pixel from e.g. the texture level map 220 to a threshold. This comparison is called a texture level criterion. The dense map is furthermore provided by comparing a depth level of a current pixel and a previous pixel, such as a depth level difference, from the depth level map 216 to a threshold. This comparison is called a depth level criterion. When both comparisons or criteria are fulfilled, the depth of the current pixel is changed to the depth of the previous pixel and the depth of the previous pixel is filled in 224 in the depth map 216. The current and previous pixels in the texture level map correspond to the current and previous pixels in the depth level map.

Alternatively or additionally to the one or more texture level criteria, the depths obtained in the sparse depth map may, possibly for each pixel or group of pixels, be accompanied with confidence levels reflecting the quality of the depth matching. These confidence levels may then be used in order to determine whether the depth of a certain pixel should be changed and filled in 224 in the depth map or not. When all the pixels in the depth map 216 have been compared with each other in accordance with the described process the dense depth map 226 is outputted, providing a densely filled depth map with objects, which have densely filled depths.

The densely filled objects are filled with depth levels that are based on the accurate depths of the edges of the objects, by taking the accurate depths of the edges and based on criteria involving both texture and depth of adjacent pixels, and by spreading out the accurate depth of the edges of the object and/or the object boundaries, by successively progressing side-wards (right- or leftwards) from an edge an repeating the above procedure.

The filling also updates the quality of the depth estimation to that of the original and well-defined pixel, which may be an edge pixel. Alternatively, the filling can also proceed in a vertical (downwards) direction once the first row of the object has been updated.

As will be explained in more detail with reference to Fig. 3, the texture level and depth level of the current and previous pixels may be calculated as e.g. an average of the pixel and pixels adjacent to the pixel. Furthermore, the depth level of the current pixel may not be changed to the depth level of the previous pixel, but may e.g. be shifted to a depth level of a pixel comprised in a local area of e.g. the previous pixel. This may be provided when e.g. a difference in an average of the texture level for the current and previous local area is relative small or e.g. below a threshold and when e.g. a difference in an average of the depth level for the current and previous local area is relative large or e.g. larger than a threshold. Even further, in such a preferred embodiment of the method according to the invention, the depth level of the current pixel may be changed to the depth level of a pixel with e.g. the lowest depth level within e.g. the previous area, as this may introduce less artifacts.

Fig. 3 shows the texture level map 302 of twelve pixels in four rows and three columns of one of the left 208 or right 210 images. These twelve pixels represent the texture level map of a part of the imaged area. In the shown example the texture level map 302 is a map of the intensity, possibly of each of the twelve pixels. The pixel with a row number (r-1) and a column number c is shown at 304, this pixel is referred to as the previous pixel. The pixel with a row number r and a column number c is shown at 306, this pixel is referred to us the current pixel.

The previous pixel local texture level local area 308 of 3x3 pixels comprises the previous pixel and pixels adjacent to the previous pixel. The current pixel local texture level local area 310 of 3x3 pixels comprises the current pixel and pixels adjacent to the current pixel. In this example, the 3x3 local pixels of the current and previous pixels intersect each other.

Fig. 3 furthermore shows the depth level map 312 of twelve pixels in four rows and three columns. The shown pixels correspond to the twelve pixels shown in the texture level map 302 and the depth level map 312 is e.g. aligned with the texture level map 302. These twelve pixels represent the depth level map of a part of the imaged area. The pixel 314 in row and column position (r-1, c) is the previous pixel and the pixel 316 in (r, c) is the current pixel in the depth map. Alternatively or additionally and eventually e.g. when starting up filling in an object, the pixel in (r, c-1) may be the previous pixel and (r, c) may be the current.

The previous pixel local depth level local area 318 of 3x3 pixels comprises the previous pixel and pixels adjacent to the previous pixel. The current pixel local depth level local area 320 of 3x3 pixels comprises the current pixel and pixels adjacent to the current pixel. The 3x3 local pixels of the current and previous pixels in this example intersect each other for a number of intersection pixels, in an intersection area 322, in this example six pixels.

The described embodiment of the method according to the invention processes the images on a row-by-row basis. A column-by-column basis or method may also be used, but here a row-by-row method is used. The reason for a row-by-row method to be used may e.g. be due to the images being provided in a row-by-row cycle or when implemented in a line based process. The result may be that the objects or pixels of the objects are filled in with changed depth levels from the left and right upper corner of the object as seen in the image. In the texture level map 302 local levels of texture are provided for the current texture level pixel 306 at (r, c) and for the previous texture level pixel 304 on the previous row (r-1, c). In the depth level map 312 an average local depth is computed or has been computed for the current depth level pixel 316 and for the previous depth level pixel 314 on the previous row. If the average local texture level of the pixel 306 and 304 are different by a threshold, e.g. smaller than a texture threshold, while the depth estimates are quite similar or even undefined or e.g. smaller than a depth threshold, then the depth level of the current pixel 316 is changed into a depth level of a pixel within the intersection area 322, and the changed depth level is filled into the depth map. Alternatively, the depth level of the current pixel 316 is changed into a depth level within the current 320 and the previous 318 local areas. In case one or more or the criteria described are not fulfilled, the depth level of the current pixel is used as a final depth level in the final dense depth map. As texture levels, the average intensity or value range or deviation or any other texture descriptor may be used. It is preferred that the depth level of the current pixel is changed into the depth level of a pixel that has the lowest depth (or lowest distance to camera or nearest to the camera or most foreground) level within a given area, such as the intersection area 322. Alternatively, it may be preferred that the depth level of the current pixel 316 is changed into the lowest depth level within the current 320 and the previous 318 local areas.

Although the present invention has been described in connection with preferred embodiments, it is not intended to be limited to the specific form set forth herein. Rather, the scope of the present invention is limited only by the accompanying claims. In this description, certain specific details of the disclosed embodiment, are set forth for purposes of explanation rather than limitation, so as to provide a clear and thorough understanding of the present invention. However, it should be understood readily by those skilled in this art, that the present invention might be practiced in other embodiments that do not conform exactly to the details set forth herein, without departing significantly from the scope of this disclosure. Further, in this context, and for the purposes of brevity and clarity, detailed descriptions of well-known apparatus, circuits and methodology have been omitted so as to avoid unnecessary detail and possible confusion. Reference signs are included in the claims, however the inclusion of the reference signs is only for clarity reasons and should not be construed as limiting the scope of the claims. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and/or by means of a suitably programmed processor. In the device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

CLAIMS:

1. A method of changing a depth level of one or more pixels, the method comprising the steps of obtaining at least one left image (208) and at least one right image (210) of an area, wherein the area is represented by a number of pixels, each pixel being adjacent to one or more other pixels, obtaining at least one texture level map (220, 222, 302) from at least one of the left or right images, obtaining a depth level map (216, 312) from the left image and the right image, changing a depth level of a pixel (306, 316) in dependence on said texture level map and/or said depth level map, and filling in (224) the changed depth level in the depth level map (216, 312).

2. The method according to claim 1, wherein said step of changing a depth level of a pixel (306, 316) in dependence on said texture level map and said depth level map comprises changing a depth level of a current pixel (306, 316) to a depth level of a previous pixel (304, 314) when one or more texture level criteria as to the current (306, 316) and previous pixels (304, 314) and/or one or more depth level criteria as to the current (306, 316) and previous pixels (304, 314) are fulfilled.

3. The method according to claim 2, wherein the one or more texture level criteria relate to the difference between the texture levels of the current (306, 316) and previous pixels (304, 314), and wherein the one or more depth level criteria between the current (306, 316) and previous pixel (304, 314) relate to the difference between the depth levels of the current (306, 316) and previous pixels (304, 314).

4. The method according to claim 1, wherein the depth level of the current pixel (306, 316) is changed to the depth level of the previous pixel (304, 314) when a difference between the texture levels of the current (306, 316) and previous pixels (304, 314) is smaller than a first threshold, and/or a difference between the depth levels of the current (306, 316) and previous (304, 314) pixels is smaller than a second threshold.

5. The method according to claim 1, wherein the depth level of a current pixel (306, 316) is changed to the depth level of a previous pixel (304, 316) when the depth level of the previous pixel (304, 314) is lower than the depth level of the current pixel (306, 316).

6. The method according to claim 1, wherein one texture level criterion and one depth level criterion are tested for local areas (308, 310, 318, 320) of the current (306, 316) and previous (304, 314) pixels, wherein the local area (308, 310, 318, 320) of the current

(306, 316) and previous (304, 314) pixels intersects for a number of intersection pixels (322), and wherein the depth level of the current pixel (306, 316) is changed to a depth level of one or more of the intersection pixels (322).

7. The method according to claim 6, wherein the depth level of the current pixel

(306, 316) is changed to a lowest depth level of one or more of the intersection pixels (322).

8. The method according to claim 6, wherein the depth level of the current pixel (306, 316) is changed when a difference in the texture level between the local areas (308, 310, 318, 320) of the current (306, 316) and the previous (304, 314) pixels is smaller than a threshold.

9. The method according to claim 6, wherein the depth level of the current pixel (306, 316) is changed when a difference in the texture level is smaller than a threshold and a difference in depth level is smaller than a threshold.

10. The method according to claim 1, wherein at least a part of an object (120, 122) within the area (118) is at least partly filled (224) with changed depth levels in order to change a relative sparse defined part (134) of the depth level map into a relative dense defined part.

11. The method according to claim 1, wherein at least an inside (134) of an object (120, 122) within the area is at least partly filled (224) with changed depth levels in a direction from a boundary (126, 128, 130, 132) of the object towards a middle (134) of a surface of the object (120, 122).

12. A control logic (214) for a depth level changing device comprising a processor programmed to carry out the method of claim 1.

13. A depth level changing system (202) comprising a control logic (214) according to claim 12, and left (204) and right side (206) cameras for obtaining at least one left image (208) and at least one right image (210).

14. A computer readable code enabling a processor to perform the method according to claim 1.