METHOD AND SYSTEM FOR DETERMINING A DEPTH IMAGE OF A SCENE

Abstract

METHOD AND SYSTEM FOR DETERMINING A DEPTH IMAGE OF A SCENE This description relates to a system (1) for determining a depth image of a scene (2), configured to project a pattern of spots (18) onto the scene and acquire an image of the scene; determine the I and Q values ​​of the pixels in the image; determine, for each pixel, at least one confidence value to form a confidence image; determine the local maxima of the confidence image whose confidence value is greater than a first threshold; select, for each local maximum, pixels around the local maximum whose confidence value is greater than a second threshold; determine a mean Imoy value equal to the average of the I values ​​of the selected pixels and a mean Qmoy value equal to the average of the Q values ​​of the selected pixels; and determine the depth of the local maximum from the mean Imoy and Qmoy values. Figure for the abstract: Fig. 2

Description

Title of the invention: METHOD AND SYSTEM FOR DETERMINING A DEPTH IMAGE OF A SCENE Technical field

[0001] The present application relates to the field of methods and systems for determining a depth image of a scene. Prior art

[0002] Image acquisition systems capable of acquiring depth information have been proposed. For example, indirect time-of-flight (iToF) detectors operate to transmit a light signal toward a scene and then detect the returning light signal reflected by objects in the scene. By evaluating the phase shift between the transmitted light signal and the reflected light signal, the distance of the scene from the image acquisition system can be estimated.

[0003] It is desirable to have a system for acquiring a depth image by indirect time of flight adapted to estimate the distance of the scene relative to the image acquisition system with high precision. Summary of the invention

[0004] There is a need to overcome all or part of the disadvantages of known indirect time-of-flight depth image acquisition systems.

[0005] One embodiment provides a method for determining a depth image of a scene, comprising the following steps: a) projecting a pattern of spots onto the scene and acquiring an image of the scene, each pixel of the image being obtained from electric charges accumulated during collection phases; b) determine the I and Q values ​​of the pixels in the image; c) determining, for each pixel of the image, at least one confidence value which is a function of at least a portion of the accumulated charges, to form a confidence image or confidence images; d) determining the local maxima of the confidence image or of one of the confidence images whose confidence value, of the same confidence image or of another of the confidence images, is greater than a first threshold; e) selecting, for each local maximum determined in step d), pixels around the local maximum whose confidence value, of the confidence image used to determine the local maxima or of another of the confidence images, is greater than a second threshold; f) determine, for each local maximum determined in step d), a value Imoy equal to the average of the I values ​​of the selected pixels and a Qmoy value equal to the average of the Q values ​​of the selected pixels; and g) determine, for each local maximum, the depth of the local maximum from the values ​​Imoy and Qmoy.

[0006] An embodiment also provides a system for determining a depth image of a scene, the system being configured to: a) projecting a pattern of spots onto the scene and acquiring an image of the scene, each pixel of the image being obtained from electric charges accumulated during collection phases; b) determine the I and Q values ​​of the pixels in the image; c) determining, for each pixel of the image, at least one confidence value which is a function of at least a portion of the accumulated charges, to form a confidence image or confidence images; d) determining the local maxima of the confidence image whose confidence value, of the same confidence image or of another of the confidence images, is greater than a first threshold; e) selecting, for each local maximum determined in step d), pixels around the local maximum whose confidence value, of the confidence image used to determine the local maxima or of another of the confidence images, is greater than a second threshold; f) determining, for each local maximum determined in step d), an Imoy value equal to the average of the I values ​​of the selected pixels and a Qmoy value equal to the average of the Q values ​​of the selected pixels; and g) determine, for each local maximum, the depth of the local maximum from the values ​​Imoy and Qmoy.

[0007] According to one embodiment, the first threshold is strictly greater than the second threshold.

[0008] According to one embodiment, step d) comprises moving a first window (F1) on the confidence image, and, for each position of the first window on the confidence image, comparing the value of a pixel of the confidence image located in the first window and which is not on the edge of the first window, to the values ​​of the other pixels of the confidence image located in the first window.

[0009] According to one embodiment, step d) comprises, in the case where the value of said pixel is greater than the values ​​of the other pixels of the confidence image located in the first window, the comparison of the value of said pixel, of the same confidence image or of another of the confidence images, with the first threshold.

[0010] According to one embodiment, step e) comprises, for each local maximum, the placing a second window on the confidence image used to determine the local maxima or on another of the confidence images, the second window containing the local maximum, and comparing to the second threshold the value of each pixel of the confidence image located in the second window, other than the local maximum.

[0011] According to one embodiment, the second window is larger than the first window or equal to the first window.

[0012] According to one embodiment, the system comprises a device for illuminating the scene with the spot pattern, and an image sensor for acquiring the image of the scene. Brief description of the drawings

[0013] These characteristics and advantages, as well as others, will be explained in detail in the following description of particular embodiments given without limitation in relation to the attached figures among which:

[0014] [Fig.l] schematically represents an embodiment of a system for acquiring a depth image by indirect time of flight;

[0015] [Fig.2] schematically represents an embodiment of a system for acquiring a depth image by indirect time of flight using a pattern of light spots;

[0016] [Fig.3] is a block diagram of one embodiment of a method of providing a depth map of a scene;

[0017] [Fig.4] is a graph illustrating an example of light intensity of a light signal emitted and returned according to one embodiment;

[0018] [Fig.5] is an example of a trust image;

[0019] [Fig.6] is an example of a confidence image on which the local maxima;

[0020] [Fig.7] is an example of a selection mask; and

[0021] [Fig.8] represents, in a partial and schematic manner, an embodiment of a processing module of the system of [Fig.l]. Description of the embodiments

[0022] The same elements have been designated by the same references in the different figures. In particular, the structural and / or functional elements common to the different embodiments may have the same references and may have identical structural, dimensional and material properties.

[0023] For the sake of clarity, only the steps and elements useful for understanding the embodiments described have been shown and are detailed. In particular, the usual electronic systems comprising an indirect time-of-flight sensor, called an iToF imaging system hereinafter, have not been detailed, the embodiments described being compatible with these usual systems.

[0024] In the following description, when reference is made to absolute position qualifiers, such as the terms "front", "back", "top", "bottom", "left", "right", etc., or relative position qualifiers, such as the terms "above", "below", "upper", "lower", etc., or to orientation qualifiers, such as the terms "horizontal", "vertical", etc., reference is made, unless otherwise specified, to the orientation of the figures or to an image in a normal viewing position.

[0025] Unless otherwise specified, the expressions "approximately", "approximately", "sen "possibly", and "of the order of" mean within 10%, preferably within 5%.

[0026] The principle of indirect time of flight consists of measuring a distance relative to an object by measuring a phase delay between an emitted light wave and a light wave reflected by the object.

[0027] [Fig.l] schematically illustrates one embodiment of an iToF imaging system 1, which can be used to provide a distance measurement of a scene 2.

[0028] The iToF imaging system 1 comprises: - a lighting device 4; - a control module 6 of the lighting device 4 configured to control the lighting device 4 to actively illuminate the scene 2 with an incident infrared light signal IL amplitude modulated by a modulation signal ILM, the incident light signal IL amplitude modulated being reflected by the scene 2 into a reflected light signal RL; - a lens 8 for collecting the reflected light signal RL forming an image of scene 2; - an image sensor 10 with a pixel matrix configured to acquire the image of the scene 2 and controlled by the control module 6, each pixel of the image sensor 10 being configured to provide an electrical signal, called pixel value hereinafter, representative of the reflected light signal RL captured by this pixel; and - a processing module 12 of the images acquired by the image sensor 10.

[0029] According to one embodiment, the image sensor 10 may comprise an electronic circuit in which the pixels are formed and the processing module 12 may also be part of this electronic circuit.

[0030] Generally, for indirect time-of-flight (iToF) operation, the processing module 12 correlates the reflected light signal RL with a demodulation signal to provide an in-phase component value ("I value") and a quadrature component value ("Q value") for each pixel, called the I and Q values. From the I and Q values ​​for each pixel, the processing module 12 determines a phase delay value ¢, also called phase shift, for each pixel, which gives a phase image. From the phase image, the processing module 12 can determine a depth value d for each pixel, which gives a depth image.

[0031] [Fig. 2] schematically illustrates the principle of distance measurement by indirect time of flight by structured type illumination. The illumination device 4 is configured to illuminate the scene 2 comprising objects 14 and 16, with a pattern of light spots, also called a pattern of light points. The image sensor 10 acquires an image of the pattern of light spots on the scene 2. The pattern of light spots projected onto the scene 2 by the illumination device 4 results in a corresponding pattern of light spots in the image acquired by the pixels of the image sensor 10.

[0032] The light spots appear in the image acquired by the image sensor 10 in the form of a spatial light pattern comprising high intensity areas 18 (the light spots themselves) and low intensity areas 20. The lighting device 4 and the image sensor 10 are spaced apart from each other by a distance B. This distance B is called the baseline. The scene 2 has a distance d from the baseline B. More precisely, each object 14, 16 or object point in the scene 2 has an individual distance d from the baseline B. The depth image of the scene provided by the iToF imaging system 1 defines a depth value for each pixel of the depth image and thus provides depth information of the scene 2 and the objects 14, 16.

[0033] Typically, the pattern of light spots 14 projected onto the scene 2 may result in a corresponding pattern of light spots captured on the pixels of the image sensor 10. In other words, spot pixel regions may be present among the plurality of pixels of the acquired image and valley pixel regions may be present among the plurality of pixels of the acquired image. The spot pixel regions of the acquired image may include signal contributions from light reflected from the scene 2 but also from background light, and / or multipath interference. The valley pixel region of the acquired image may include signal contributions from background light and / or multipath interference.

[0034] For example, the lighting device 4 may produce from 200 to 2000 spots 18 on the scene 2, for example about 1000 spots 18 on the scene 2. The light spots may have a circle or rectangle / square shape or any other regular or irregular shape. The light pattern of spots may be a grid pattern or a line pattern or an irregular pattern. Preferably, the lighting device 4 is configured so that each spot 18 has a maximum intensity at its center.

[0035] [Fig. 3] is a block diagram of one embodiment of a method for deter- mination of a depth image implemented by the iToF imaging system.

[0036] In step 30 (Image), raw image data is received by the image sensor 10.

[0037] [Fig. 4] is a graph representing, by a curve C_IL, an example of evolution as a function of time of the light intensity of the incident light signal IL modulated in amplitude emitted by the lighting device 4 towards the scene 2, and, by a curve C_RL, an example of evolution as a function of time of the light intensity of the reflected light signal RL received by one of the pixels of the image sensor 10. Although, to simplify the comparison, these signals are represented in [Fig. 4] as having substantially the same intensity, in practice the reflected light signal RL received by each pixel is likely to be notably less intense than the incident light signal IL. In the example of [Fig. 4], the incident light signal IL is modulated in amplitude by a modulation signal ILM corresponding to a sinusoidal signal of period T and frequency fmod.Each of the light intensity of the incident light signal IL and the light intensity of the reflected light signal RL therefore has the form of a sinusoidal wave of period T. However, in alternative embodiments, it could have a different periodic form, for example consisting of a sum of sinusoidal waves, of triangular shape, or in square waves. As can be seen in [Fig.4], the received light signal RL has a non-zero average Offset value and a maximum amplitude Amp relative to this average value.

[0038] The phase shift <t>is for example estimated from a sampling, for each pixel of the image sensor 10, of the reflected light signal RL captured by the pixel during at least three distinct sampling windows, preferably during four distinct sampling windows, during each period T of the reflected light signal RL. As an example, in [Fig.4], the acquisition of four samples per period has been illustrated.

[0039] The samples of each sampling window are for example integrated over a large number of periods for example over approximately 100,000 periods, or more generally between 10,000 and 10 million periods. Each sampling window has for example a duration of up to a quarter of the period of the light signal. These sampling windows are named C1, C2, C3, and C4 in [Fig. 4], and in the example of [Fig. 4], each sampling window is of the same duration and the four sampling windows have a total cycle time equal to the period of the light signal. More generally, there may or may not be a time interval separating one sampling window from the next and, in some cases, there could be an overlap between the sampling windows.Each sampling window has, for example, a duration between 15% and 35% of the period of the light signal in the case of a pixel capturing four samples per period. Each sampling window has, for example, a duration of between 25% and 40% of the period of the light signal in the case of a pixel capturing three samples per period. The timing of the sampling windows C1 to C4 is controlled so as to be synchronized with the timing of the incident light signal IL.

[0040] According to one embodiment, the sampling window C1 is phase-shifted by 0° relative to the modulation signal ILM. During the sampling window C1, each pixel of the image sensor 10 accumulates an electric charge Q1 which depends on the quantity of light received by the pixel during the sampling window C1. The sampling window C2 is phase-shifted by 90° relative to the modulation signal ILM. During the sampling window C2, each pixel of the image sensor 10 accumulates an electric charge Q2 which depends on the quantity of light received by the pixel during the sampling window C2. The sampling window C3 is phase-shifted by 180° relative to the modulation signal ILM. During the sampling window C3, each pixel of the image sensor 10 accumulates an electric charge Q3 depending on a quantity of light received by the pixel during the sampling window C3.The sampling window C4 is 270° out of phase with the modulation signal ILM. During the sampling window C4, each pixel of the image sensor 10 accumulates an electrical charge Q4 which depends on the amount of light received by the pixel during the sampling window C4. The method continues at step 32.

[0041] The determination of the electric charges Q1, Q2, Q3, and Q4 can be carried out according to a sampling method other than that described previously. Another technique based on the detection of four samples per period is described in more detail in the publication by R. Lange and P. Seitz entitled "Solid-state TOF range camera", IEEE J. on Quantum Electronics, vol. 37, No. 3, Tuesday 2001, the contents of which will be considered as an integral part of the present description to the extent permitted by law.

[0042] In step 32 (I, Q), the values ​​I and Q are determined for each pixel from the raw image data provided by the image sensor 10.

[0043] According to one embodiment, the quadrature component Q is given by the following relation: [Math 1] Q=Q3-Q4

[0044] According to one embodiment, the phase I component is given by the following relation: [Math 2] I=Q1-Q2

[0045] The method continues at step 34.

[0046] In step 34 (Conf), a confidence map, also called a confidence image, is determined. According to one embodiment, a confidence value conf is determined for each pixel from the I and Q values ​​of the pixel. Generally, the confidence value conf is a function of both the I value and the Q value of the pixel.

[0047] [Fig.5] is an example of a C_conf confidence map obtained in step 34.

[0048] According to one embodiment, the confidence value Conf corresponds to the amplitude AmpLl said according to the L1 norm of the reflected light signal RL and is given by the following relation: [Math 3] Conf= AmpL 1 =^I 2 +Q 2

[0049] According to another embodiment, the confidence value Conf corresponds to the amplitude AmpL2 said according to the L2 norm of the reflected light signal RL and is given by the following relation: [Math 4] conf=AmpL2= 11 j+1Q |

[0050] According to another embodiment, the confidence value Conf corresponds to a function of the amplitude Amp, corresponding to AmpL1 or AmpL2, and of the Offset value of the pixel, and is for example equal to the square of the signal-to-noise ratio SNR, for example according to the following relationship: [Math 5] CQnf_Amp2 where the Offset value is for example given by the relation k Offset following in the case of four samples per period: [Math 6] Offset=Q 1+Q2+Q3+Q4

[0051] As a variant, in the case where the sampling is carried out by half-period, the Offset value is equal to the sum of Q1 and Q3 and is also equal to the sum of Q2 and Q4.

[0052] According to another embodiment, the confidence value Conf corresponds to any other estimator of the signal-to-noise ratio SNR of the pixel.

[0053] The method continues at step 36.

[0054] In step 36 (Max), a method for determining local maxima of the confidence map is implemented and is illustrated by [Fig.6]. According to one embodiment, the method consists of sliding, on the confidence map, a first window F1, the outline of which is indicated in dotted lines in [Fig.6], onto the confidence map C_conf, and comparing the value of a pixel, called the tested pixel, of the confidence map located in the first window F1 and which is not on the edge of the first window F1, to the values ​​of the other pixels of the confidence map located in the first window Fl. According to one embodiment, the first window Fl corresponds to a group of contiguous pixels. According to one embodiment, the pixel tested is the pixel in the center of the first window Fl. According to one embodiment, the first window Fl is a square window. Generally, the size of the first window Fl depends on the distance between the spots 18. According to one embodiment, the first window is a square window with sides of 3 to 7 pixels, preferably 5 pixels.

[0055] If the value of the tested pixel is not greater than all the values ​​of the other pixels of the first window F1, the first window F1 is moved to the next position. If the value of the tested pixel is greater than all the values ​​of the other pixels of the first window F1, and if the value of the tested pixel is greater than a first threshold, the tested pixel of the first window F1 is considered as a local maximum of the confidence map C_conf and the position of this local maximum in the confidence map C_conf is stored. The first window F1 is moved to the next position.

[0056] Step 36 is completed when the first window F1 has been moved on the trust map C_conf to cover the entire trust map C_conf. The method continues at step 38.

[0057] [Fig.7] represents the confidence map C_Conf of [Fig.5] on which each local maximum MaxL obtained in step 36 is represented by a white pixel with a cross.

[0058] In step 38 (Selection), a method for selecting pixels from the confidence map C_conf is implemented for each local maximum MaxL determined in step 36 and is illustrated by [Fig.8]. According to one embodiment, for each local maximum MaxL, a step for selecting pixels from the confidence map C_conf neighboring the local maximum MaxL is implemented. According to one embodiment, for each local maximum MaxL, a second window F2 is applied to the confidence map, the outline of which is indicated in dotted lines in [Fig.8], containing the local maximum MaxL, the local maximum not being on the edge of the second window F2, and the value of each pixel of the confidence map located in the second window F2, other than the local maximum MaxL, is compared to a second threshold. According to one embodiment, the first threshold and the second threshold are determined by tests. The second threshold is strictly lower than the first threshold.According to one embodiment, the second window F2 corresponds to a group of contiguous pixels of the confidence map. According to one embodiment, the second window F2 is larger than the first window FL. According to one embodiment, the second window F2 is centered on the local maximum MaxL. According to one embodiment, the second window F2 is a square window. Generally, as for the first window Fl, the size of the second window F2 depends on the distance between the spots 18. According to one embodiment. embodiment, the first window is a square window with sides of 1 to 31 pixels, preferably 5 to 9 pixels, more preferably 7 pixels. For each local maximum MaxL, the set of pixels comprising the local maximum MaxL and the pixels of the second window F2 which are selected in step 38 form a confidence zone. The positions of the pixels in each confidence zone are stored. The method continues in step 40.

[0059] [Fig.9] represents a Mask image of the same dimension as the confidence map of [Fig.5] and on which each confidence zone ZC determined in step 38 is represented by white pixels, the other pixels of the image being black.

[0060] In the embodiment described above, the same confidence map is used in step 36 for the comparison to the first threshold, in step 36 for the determination of the local maxima, and in step 38 for the comparison to the second threshold. According to another embodiment, different confidence maps may be used in step 36 for the comparison to the first threshold, in step 36 for the determination of the local maxima, and / or in step 38 for the comparison to the second threshold. Preferably, the same confidence map is used in step 36 for the comparison to the first threshold and in step 38 for the comparison to the second threshold.For example, the confidence map obtained according to the relation Math 3 can be used in step 36 for the comparison at the first threshold, the confidence map obtained according to the relation Math 5 can be used in step 36 for the determination of the local maxima, and the confidence map obtained according to the relation Math 3 can be used in step 38 for the comparison at the second threshold.

[0061] In step 40 (Signal Aggregation), a correction of the I and Q values ​​determined in step 32 is implemented. For each confidence zone determined in step 38, the processing module 12 determines the average Imoy of the I values ​​of the pixels in the confidence zone and determines the average Qmoy of the Q values ​​of the pixels in the confidence zone. The method continues in step 42.

[0062] In step 42 (¢), a phase delay O is determined for each pixel of each confidence zone on the basis of the values ​​Imoy and Qmoy. According to one embodiment, the phase delay value O is given by the following relationship: [Math 8] 0=arctan(^7)

[0063] The method continues at step 44.

[0064] In step 44(d), a depth value d is determined for each confidence zone based on the phase delay value O of the confidence zone. According to one embodiment, the distance d to the object is given by the following relationship: [Math 9] ([=— ...... <j)où c est la vitesse de lumière. 2jt 2fmod

[0065] The distance d determined for each pixel provides the depth image containing spots 50, the color of each spot 50 depending on the depth d determined in step 44 for the spot considered.

[0066] [Fig. 10] is an example of an lmp depth image.

[0067] In previously described embodiments, method steps may be performed using one or more computing devices. The embodiments are not limited to operation with a particular type of computing device.

[0068] [Fig. 11] is a block diagram of a computing device 1000 that may be used to implement the processing module 12 and / or the control module 6. The computing device 1000 may include one or more processors 1001 (Processor(s)) and one or more non-transitory computer-readable storage media (e.g., memory 1003 (Memory)). The memory 1003 may store, in a non-transitory computer-readable storage medium, computer program instructions that, when executed, implement the steps of the detection method described above. The processor(s) 1001 may be coupled to the memory 1003 and may execute these computer program instructions to cause these steps to be performed.

[0069] The computing device 1000 may also include a Network FO Interface(s) 1005 through which the computing device may communicate with other computing devices (e.g., over a network), and may also include one or more User FO Interface(s) 1007 through which the computing device may provide an output signal to a user and receive an input signal from the user. The user interfaces may include peripherals such as a keyboard, a mouse, a microphone, a display device (e.g., a monitor or touchscreen), speakers, a camera, and / or various other types of input / output devices.

[0070] The embodiments described above may be implemented in several ways. For example, the embodiments may be implemented using dedicated circuitry, software, or a combination thereof. When implemented by software, the software code may be executed on any suitable processor (e.g., a microprocessor) or set of processors, whether provided in a single computing device or distributed among multiple computing devices. It should be noted that any component or set of components that perform the method steps described above may be considered one or more several controllers that control the steps described above. The controller(s) may be implemented in many ways, for example with a dedicated electronic circuit or with a general-purpose circuit (e.g., one or more processors) that is programmed using software or microcode to perform the process steps described above.

[0071] In this regard, it should be noted that one embodiment described herein includes at least one computer-readable storage medium (RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical disc medium, magnetic cassette, magnetic tape, magnetic storage disc or other magnetic storage device, or other non-transitory computer-readable storage medium) encoded with a computer program (i.e., multiple executable instructions) that, when executed on one or more processors, performs the steps of the embodiments described above. The computer-readable medium may be transportable such that the program stored thereon may be loaded onto any computing device to implement aspects of the techniques described herein.Furthermore, it should be noted that reference to a computer program that, when executed, performs one of the method steps described above, is not limited to an application program running on a host computer. Rather, the terms computer program and software are used herein in a general sense to refer to any type of computer code (e.g., application software, firmware, microcode, or any other form of computer instruction) that can be used to program one or more processors to implement aspects of the methods described above.

[0072] Various embodiments and variations have been described. Those skilled in the art will understand that certain features of these various embodiments and variations could be combined, and other variations will occur to those skilled in the art. Finally, the practical implementation of the described embodiments and variations is within the ability of those skilled in the art from the functional indications given above. < / t>

Claims

Claims

1. A method for determining a depth image (lmp) of a scene (2), comprising the following steps: a) projecting a pattern of spots (18) onto the scene and acquiring an image of the scene, each pixel of the image being obtained from electrical charges accumulated during collection phases; b) determining the phase component values ​​I and the quadrature component values ​​Q of the pixels of the image; c) determining, for each pixel of the image, at least one confidence value which is a function of at least a portion of the accumulated charges, to form a confidence image or confidence images; d) determining the local maxima of the confidence image or of one of the confidence images whose confidence value, of the same confidence image or of another of the confidence images, is greater than a first threshold;e) selecting, for each local maximum determined in step d), pixels around the local maximum whose confidence value, from the confidence image used to determine the local maxima or from another of the confidence images, is greater than a second threshold; f) determining, for each local maximum determined in step d), an Imoy value equal to the average of the I values ​​of the selected pixels and a Qmoy value equal to the average of the Q values ​​of the selected pixels; and g) determining, for each local maximum, the depth of the local maximum from the Imoy and Qmoy values.;

2. The method of claim 1, wherein the first threshold is strictly greater than the second threshold.

3. A method according to claim 1 or 2, wherein step d) comprises moving a first window (F1) on the confidence image, and, for each position of the first window on the confidence image, comparing the value of a pixel of the confidence image located in the first window and which is not on the edge of the first window, to the values ​​of the other pixels of the confidence image located in the first window.

4. A method according to claim 3, wherein step d) comprises, in the case where the value of said pixel is greater than the values ​​of the others pixels of the confidence image located in the first window, comparing the value of said pixel, of the same confidence image or of another of the confidence images, to the first threshold.

5. Method according to any one of claims 1 to 4, in which step e) comprises, for each local maximum, the placement of a second window (F2) on the confidence image used to determine the local maxima or on another of the confidence images, the second window containing the local maximum, and the comparison with the second threshold of the value of each pixel of the confidence image located in the second window, other than the local maximum.

6. The method of claims 3 and 5, wherein the second window (F2) is larger than the first window (F1) or equal to the first window (F1).

7. System (1) for determining a depth image (lmp) of a scene (2), the system being configured to: a) project a pattern of spots (18) onto the scene and acquire an image of the scene, each pixel of the image being obtained from electrical charges accumulated during collection phases; b) determine the phase component values ​​I and the quadrature component values ​​Q of the pixels of the image; c) determine, for each pixel of the image, at least one confidence value which is a function of at least a portion of the accumulated charges, to form a confidence image or confidence images; d) determine the local maxima of the confidence image whose confidence value, of the same confidence image or of another of the confidence images, is greater than a first threshold;e) selecting, for each local maximum determined in step d), pixels around the local maximum whose confidence value, from the confidence image used to determine the local maxima or from another of the confidence images, is greater than a second threshold; f) determining, for each local maximum determined in step d), an Imoy value equal to the average of the I values ​​of the selected pixels and a Qmoy value equal to the average of the Q values ​​of the selected pixels; and g) determining, for each local maximum, the depth of the local maximum from the Imoy and Qmoy values.;

8. A system according to claim 7, comprising a device illumination (4) of the scene (2) with the spot pattern (18), and an image sensor (10) for acquiring the image of the scene.

Images