Glare reduction in images

Abstract

An example non-transitory machine-readable medium comprising instructions for capturing a first image of a scene comprising light emitted by a display device, changing a brightness of the display device, capturing a second image of the scene while the brightness of the display device is changing, training a machine learning model with the first image and the second image to provide a light filter to reduce glare, and applying the machine learning model to a captured third image of the scene to reduce glare in the third image, the third image being different from the first image and the second image.

Description

Background Technology

[0001] Video capture typically involves capturing time-series image frames. It can be used in video conferencing to provide visual communication between various users in different locations over a computer network. Video conferencing can be facilitated by real-time video capture performed by computing devices at different locations. Video capture can also be used in other applications, such as recording video for later playback. Attached Figure Description

[0002] Figure 1 It is a block diagram of an exemplary non-transitory machine-readable medium including glare reduction instructions that control a light source to train a machine learning model to remove or reduce glare in a captured image.

[0003] Figure 2 This is a flowchart of an exemplary method for controlling a light source to train a machine learning model to remove or reduce glare in a captured image.

[0004] Figure 3 This is a flowchart of an exemplary method for controlling a light source to train a machine learning model to remove or reduce glare in a captured image, including training the machine learning model in response to an event.

[0005] Figure 4 This is a block diagram of an exemplary device for controlling a light source to train a machine learning model to remove or reduce glare in a captured image.

[0006] Figure 5 This is a block diagram of an exemplary device for controlling light sources to train a machine learning model to remove or reduce glare in a captured image, wherein such glare is caused by multiple light sources. Detailed Implementation

[0007] Images captured by digital video frames can include glare caused by an object's glasses or other reflective surfaces such as badges, clear masks, faceplates, metal badges, fashion accessories, etc. During a video conference, this glare can be generated by light emitted from a participant's display device and captured by the participant's camera. Glare in a video conference moves and changes as the participant moves their head in three dimensions (e.g., xyz translation, yaw, pitch, and roll) and as the content on their display device changes. This can distract other users in the video conference and may reduce the realism of the meeting by subtly reminding participants that they are communicating via cameras and display devices. Furthermore, glare can reduce privacy and confidentiality because sensitive information (e.g., document pages) may be visible in the reflections. Even if the content in the glare reflections is difficult to understand or unreadable, discernible characteristics of the glare, such as color, shape, and movement, can still reveal sensitive information.

[0008] Glare caused by light with transmission properties reflected from eyeglasses or other movable reflective surfaces can confuse simple filters. Furthermore, in video conferencing, this glare often cannot be reduced simply by moving the light source causing it, as the position of the light source is usually conducive to the smooth operation of the video conference.

[0009] The brightness of a display device can be modulated to alter glare in captured images. Such images can be used to train machine learning models to provide filters for glare removal. For example, the display backlight can be briefly turned off ("blanking") to prevent display glare in video frames, after which the backlight returns to its normal brightness. In other examples, the backlight brightness can be increased or maximized. This captures frames with different levels of glare. The machine learning model calculates filters for glare removal based on information provided by such captured frames. That is, images of the same scene that are temporally close and have different brightness levels and generated glare provide a representation of glare to train the machine learning model. The trained model can be applied to newly captured frames to reduce or eliminate glare. Alternatively, other brightness levels besides "blanking" can also be used to quantify and train models for glare removal. For example, by using multiple brightness levels, complete glare can be avoided, and in doing so, the effect of glare can be reduced (which can reduce the total brightness in a way that is perceptible to the user).

[0010] Since glare can move and change its characteristics during a video conference, target frames with reduced glare can be captured at intervals, allowing machine learning models to be trained continuously during the conference. The hidden-face removal rate can decrease over time, enabling higher hidden-face removal rates in the initial phase of camera activity for model training, and lower rates in later phases of camera activity.

[0011] These same technologies can be used in other video capture applications, such as capturing video for later playback.

[0012] Figure 1 An exemplary non-transitory machine-readable medium 100 is shown, including glare reduction instructions 102 that remove or reduce unwanted glare in captured images. The glare reduction instructions can implement a dynamic filter as discussed below, allowing glare to be removed or reduced in real-time or near real-time, such as during live video conferencing or during another type of video capture. This reduces or eliminates viewer distraction or perceived quality degradation that may be caused by glare (such as glare typically caused by eyeglasses).

[0013] The non-transitory machine-readable medium 100 may include electronic, magnetic, optical, or other physical storage devices that encode instructions. This medium may include, for example, random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, memory drives, optical components, etc.

[0014] The medium 100 can work with a processor, which may include a central processing unit (CPU), microcontroller, microprocessor, processing core, field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), or similar device capable of executing instructions.

[0015] Glare reduction instructions 102 can be executed directly, such as in binary files, and / or may include interpretable code, bytecode, source code, or similar instructions that may undergo additional processing before execution.

[0016] Instruction 102 captures a first image 104 of a scene 106 including light 108 emitted by display device 110. The first image 104 may be a video frame. The image capture can be performed using a camera, such as a webcam used during video conferencing or other video capture processes. The first image is expected to include glare.

[0017] Display device 110 may be a monitor used during video conferencing or video capture. Display device 110 may include a liquid crystal display (LCD), a light-emitting diode (LED) display, or a similar display device. Display device 110 may have controllable brightness, such as controllable backlight.

[0018] The camera and display device 110 may be directed at the same user, who is part of the scene and may be a participant in a video conference or may otherwise capture their own video. The light emitted by display device 110 108 may cause glare in the captured image, for example, by reflection from the user's glasses. In other examples, another light source causes glare, such as a lamp (e.g., a ring light). Thus, the first image 104 is expected to include glare.

[0019] Instruction 102 changes the brightness of display device 110 or other light source causing glare, and then captures a second image 112 of scene 106 while the brightness is reduced. The change in display brightness can be achieved, for example, by temporarily turning off the backlight of display device 110 for one frame of video capture. Turning off the backlight can be referred to as blanking the display. The brightness of display device 110 can be reduced or blanked for any appropriate duration, which can be quantified by a number of frames, such as one, two, or three. The shorter the blanking duration, the less likely the blanking will be noticeable to the user, who is typically expected to look at display device 110 during a video conference. Because display device 110 is temporarily turned off during the capture of the second image 112, the second image 112 does not include significant glare caused by display device 110. The same can be said of another controllable light source, such as a light that a user can use to illuminate their face during a video conference.

[0020] In other examples, instead of being reduced or blanked, or in addition to being reduced or blanked, the brightness of display device 110 is temporarily increased or maximized. The temporarily increased brightness can temporarily increase glare, and this information, together with an image having normal glare, is sufficient to identify and characterize the glare in the normal state that is expected to be removed. While the examples discussed herein consider temporarily reducing the brightness of the display device to obtain an image with reduced glare, it should be understood that brightness can also be additionally or alternatively temporarily increased to obtain an image with increased glare to achieve a comparable result.

[0021] The first image 104 is a true-brightness image of the scene 106 forming a video conference or video, or a similar image forming a video conference or video, while the second image 112 is a reduced-brightness image used for glare correction. The terms "first," "second," "third," etc., do not restrict the temporal order of image capture. For example, the first image 104 may be captured before or after the second image 112. The accuracy of glare correction increases as the first and second images 104, 112 are closer together in time.

[0022] Instruction 102 trains a machine learning model (ML model) 114 using the first and second images 104, 112. Since the first and second images 104, 112 are temporally close (e.g., within 1-3 video frames), they represent approximately the same physical representation of scene 106. That is, the differences between the first image 104 and the second image 112 caused by motion in scene 106 may be small. This is especially true in video conferencing where objects in scene 106 typically do not move very quickly. Therefore, the first and second images 104, 112 can be considered to represent two versions of the same scene: 1) a version with true brightness (first image 104) and 2) a reduced brightness version (second image 112) without glare caused by display device 110. The first image 104 has a normal overall brightness level and may contain glare. The second image 112 has a reduced overall brightness level and reduced or eliminated glare. Therefore, the machine learning model 114 is provided with sufficient information to characterize the glare caused by display device 110. In this way, machine learning model 114 can be trained to provide filters to reduce such glare.

[0023] Machine learning model 114 may include convolutional neural networks (CNNs), such as augmented causal CNNs. Augmented causal CNNs can be configured as revisionists, where data can be fed back to help re-evaluate past data samples.

[0024] In various examples, a second image 112 is provided as a brightness target for a machine learning model 114. The model 114 is then trained to generate filters that make the first image 104 approximate the brightness target. Conceptually, the second image 112 can be considered a two-dimensional map of the target brightness level, and the machine learning model 114 can be trained to filter the first image 104 to conform to this map as closely as possible.

[0025] Brightness can be an intensity independent of color, because the content displayed on display device 110 and the resulting glare can contain a variety of colors. Machine learning model 114 can be trained to filter glare regardless of its color composition.

[0026] Instruction 102 can capture a reduced brightness (second) image 112 and train a machine learning model 114 at various intervals to continuously train the machine learning model 114 during a video conference or video capture. For example, during a video conference, the capture of the reduced brightness image 112 and the training of the machine learning model 114 can be performed every 30, 60, or 90 frames. Capturing a true brightness (first) image 104 is incidental, as these are the images that constitute the captured video. The reduced brightness image 112 can be omitted from the captured video and discarded after being used to train the model 114. A temporally close true brightness image 104 can be copied to replace the omitted reduced brightness image 112.

[0027] Instruction 102 applies machine learning model 114 to a captured third image 116 of scene 106 to reduce glare in the third image 116. The third image 116 is a true brightness image different from the first and second images 104, 112. For example, machine learning model 114 can be applied to a sequence of video frames (third image 116) between capture and training intervals using the reduced-brightness image 112 to filter the captured video to remove or reduce glare. All or most of the video can be formed from the third image 116 filtered using the trained machine learning model 114. The first image 104 can also be filtered to be included in the video. The second image 112 can be discarded.

[0028] Training the machine learning model 114 may take time and does not need to be completed immediately after capturing the first and second images 104, 112. The aforementioned third image 116 can occur several frames, seconds, or minutes after capturing the first and second images 104, 112. Training can be initiated shortly after capturing the first and second images 104, 112 and can be permitted to occur based on other constraints, such as available processing and memory resources not used for video capture. Simultaneously, an earlier version of the trained machine learning model 114 can be used. Therefore, a copy of the machine learning model 114 can be trained while the original is being used to filter glare. When training is complete, the copy becomes the new original, and a new copy can be made during the next training session.

[0029] Instruction 102 can control the frequency of intervals for capturing the (second) brightness image 112 and training the machine learning model 114. The frequency can be controlled based on the error function of model 114 or based on the content displayed by display device 110.

[0030] Instruction 102 can apply an error (or loss) function when applying machine learning model 114. For example, backpropagation can be performed using a glare-corrected third image 116. The error function can be used to control the frequency of capturing the (second) image 112 with reduced brightness. Larger errors may increase the frequency. For example, during a video conference, a sudden change in user posture or orientation may increase the error. In response, instruction 102 can increase the frequency of capturing the reduced-brightness image 112 and training the model to dynamically react to changes in glare that increase the error. Conversely, when training model 114 during the video conference process, the frequency of capturing the reduced-brightness image 112 and training the model can be decreased as the error decreases due to its increased accuracy.

[0031] Instruction 102 can trigger a reduction in the brightness of display device 110 and the capture of a reduced-brightness image 112 based on the content displayed in the video conference. That is, the content displayed on display device 110 can change over time and can be used as an interval to trigger the training of machine learning model 114. For example, when the content changes significantly (e.g., from the faces of video conference participants to a shared document), a reduced-brightness image 112 can be captured, and machine learning model 114 can be trained to account for possible changes in glare corresponding to the change in content.

[0032] As described above, glare reduction instruction 102 provides real-time or near real-time correction for glare in captured video, such as during a video conference. As the machine learning model 114 is trained during capture, the glare correction becomes progressively more accurate. Furthermore, glare correction can dynamically respond to changes in the video, such as those caused by object movement and content presentation (e.g., screen sharing).

[0033] Figure 2 An exemplary method 200 for reducing glare in captured images such as video frames is shown. Method 200 can be implemented using instructions that can be stored in a non-transitory machine-readable medium and executed by a processor. Details of the elements of method 200 described elsewhere herein will not be repeated in detail below; related descriptions provided elsewhere herein may be referenced to elements identified by similar terms or reference numerals.

[0034] At frame 202, a first image of the scene is captured. The first image includes light emitted by a light source such as a display device, lamp, or other controllable light source. The display device may be used to assist in video conferencing. Users may use lamps to illuminate their faces or other objects during video conferencing or video capture. Any suitable combination of controllable light sources, such as multiple monitors, may be used. Glare may occur in the first image due to the user's glasses or other reflective surfaces.

[0035] At box 204, the light source is controlled to output a changed light intensity, such as a reduced (e.g., blanking) intensity or an increased (e.g., maximized) intensity. In the example of a display device, the backlight can be temporarily turned off or blanked. In the example of multiple display devices, for a given execution of box 204, one display device can be turned off. The controllable lamp can be temporarily turned off or dimmed. In other examples, the brightness of the display or lamp can be set to its highest setting. The amount of time for which the light output is changed can be selected to be sufficient to capture an image or a video frame. For example, the light source can be controlled to have a reduced output within a frame, or a reduced output of approximately 1 / 30 of a second when capturing video at 30 frames per second (FPS).

[0036] At box 206, a second image is captured from the scene illuminated by varying light intensity. The second image will have glare characteristics different from the light source. The light source is intentionally modulated to make the second image less or more affected by glare. The second image is captured temporally close to the first image. For example, the second image may be captured immediately before or after the first image (e.g., approximately 1 / 30th of a second before or after the first image in a 30FPS video). In another example, the second image is captured two or three frames before or after the first image (e.g., approximately 1 / 15 to 1 / 10th of a second before or after the first image in a 30FPS video). Other temporal proximity is also appropriate, given the understanding that the closer the first and second images are temporally, the smaller the motion or other differences between the first and second images that affect glare correction.

[0037] At box 208, a machine learning model is trained using the first and second images. The first and second images represent a scene with glare under normal lighting and a scene under reduced / increased lighting and reduced / increased glare, respectively. This information is sufficient to characterize glare, thereby training the machine learning model to filter out glare for subsequent images of the same scene. An example of a suitable machine learning model is given above.

[0038] At box 210, a machine learning model is applied to a captured third image of the scene to reduce glare in the third image. The third image can be captured after the machine learning model has been trained based on the first and second images. The glare-filtered third image can be included in the video capture. The trained model can be applied to any suitable number of third images.

[0039] During a video conference, a user's camera can capture first, second, and third images to correct glare caused by light sources, such as the user's display device also used in the video conference, that emit light reflected from the user's glasses or other surfaces in the scene. The first and second images can be captured intermittently to train a machine learning model. The third image can be captured sequentially and processed by the machine learning model to create a video with reduced glare.

[0040] Method 200 can be repeated continuously via frame 212 during the duration of a video conference or other video capture.

[0041] Figure 3 An exemplary method 300 for reducing glare in captured images, such as video frames, is shown, comprising training a machine learning model in response to events such as changes in an error function or content. Method 300 can be implemented using instructions that can be stored in a non-transitory machine-readable medium and executed by a processor. Details of the elements of method 300 described elsewhere herein will not be repeated in detail below; related descriptions provided elsewhere herein may be referenced to elements identified by similar terms or reference numerals.

[0042] At box 202, capture a true brightness image of the scene using a lighting source that can cause glare. The true brightness image may contain unwanted glare.

[0043] At box 210, the trained machine learning model is applied to the real brightness image to obtain an image with reduced glare.

[0044] At box 302, the glare-reduced image is then output as a frame in the video conference or otherwise provided as part of the captured frames in the video. The output of video frames may include displaying locally captured frames, transmitting frames over a computer network for remote display, saving frames to local storage, or a combination thereof.

[0045] At box 304, it is determined whether an event has occurred that triggered the training of the machine learning model. An example of a suitable event is an error in an image where glare reduction exceeds an acceptable error. That is, the error (or loss) in the glare-reduced image can be calculated and compared to an acceptable error. If the error is unacceptable, an error event occurs. Another suitable exemplary event is a change in content at the display device that acts as a light source generating glare in a true-brightness image. If the content generating glare changes, the characteristics of the glare can also change. Thus, a content event can be said to have occurred.

[0046] If the event has not yet occurred, then repeat frames 202, 210, 302, and 304 for the next frame. Therefore, the video can be continuously corrected for glare.

[0047] If the event occurs, the machine learning model is trained via boxes 204, 206, and 208. The light source causing glare temporarily alters its output (box 204), allowing the capture of an image with altered glare (box 206). The image with altered glare and a temporally close image of the true brightness are then used to train the machine learning model (box 208). Method 300 continues with boxes 202, 210, 302, and 304 to correct for glare in subsequently captured images.

[0048] Figure 4 An exemplary device 400 for removing or reducing glare in captured images is shown. Details of the elements of the device 400 described elsewhere herein will not be repeated in detail below; related descriptions provided elsewhere herein may be referenced to elements identified by similar terms or reference numerals.

[0049] Device 400 can be a computing device, such as a laptop computer, desktop computer, all-in-one (AIO) computer, smartphone, tablet computer, etc. Device 400 can be used to capture video in situations such as video conferencing, and such video may be subject to glare caused by light emitted by components of device 400.

[0050] Device 400 includes a light source such as display device 402, camera 404, and processor 406 connected to display device 402 and camera 404. In addition to or in place of display device 402, the light source may also include lamps or similar controllable light sources.

[0051] In this example, display device 402 includes a backlight 408. Display device 402 displays content 410 that may be related to video capture or video conferencing. Content 410 may include images of a teleconference scenario remote from device 400, shared documents, collaborative whiteboards, etc.

[0052] Camera 404 may include a network camera or similar digital camera capable of capturing video.

[0053] Display device 402 or other light sources and camera 404 may be directed to user 412 of device 400.

[0054] Examples of suitable processors 406 have been discussed above. As mentioned above, a non-transitory machine-readable medium 414 may be provided for operation in conjunction with the processor.

[0055] Device 400 also includes a machine learning model 416 that provides a glare reduction filter to the video 418 captured by camera 404. An example of a suitable machine learning model 416 is given above.

[0056] Device 400 may also include a network interface 420 to provide data communication for video conferencing. Network interface 420 includes hardware such as a network adapter card, network interface controller, or network-enabled chipset, and may also include instructions such as drivers and / or firmware. Network interface 420 allows data to be transmitted via computer network 422, such as a local area network (LAN), wide area network (WAN), virtual private network (VPN), the Internet, or a similar network that may include wired and / or wireless paths. Communication between device 400 and other devices 400 may be performed via computer network 422 and the corresponding network interface 420 of such device 400.

[0057] The device 400 may also include a video capture application 424, such as a video conferencing application. The application 424 may be executed by the processor 406.

[0058] The processor 406 controls the camera 404 to capture a sequence 426 of video frames or images, such video frames can be used by an application to provide video conferencing.

[0059] During normal image capture, a light source, such as display device 402, may illuminate user 412 of device 400, either intentionally as in the example of a lamp or as a side effect as in the example of display device 402. This illumination may cause glare through, for example, the user's eyes. Processor 406 applies machine learning model 416 to the captured images 428 in sequence 426 to reduce such glare.

[0060] The processor 406 further reduces the intensity of the light source during the capture of the target, dimmed image 430 used to train the machine learning model 416. This can be done by temporarily turning off the backlight 408 of the display device 402. The target image 430 can be captured at intervals 432, such as in response to excessive errors (losses) in the machine learning model 416 or triggered by changes in the content 410 at the display device 402, which acts as the light source.

[0061] Processor 406 trains machine learning model 416 using target image 430 and another normal brightness image 428 in sequence 426 that is temporally close to target image 430. The range of brightness information provided by target image 430 and normal brightness image 428 is sufficient to train machine learning model 416 to filter glare from other images 428 in sequence 426.

[0062] After the trained instances, the processor 406 continues to apply the machine learning model 416 to subsequent images 428 in sequence 426 in order to reduce glare in subsequent images 428.

[0063] Training can be performed at intervals 432 during video capture, and the machine learning model 416 can thus filter glare more accurately as the object 412 captured by the camera 404 moves and the characteristics of the light emitted by the light source change over time.

[0064] Figure 5 An exemplary device 500 for removing or reducing glare in a captured image is shown, wherein such glare may be caused by multiple light sources. Details of the elements of device 500 described elsewhere herein will not be repeated in detail below; related descriptions provided elsewhere herein may be referenced to elements identified by similar terms or reference numerals. Except as discussed below, device 500 is similar to device 400.

[0065] Device 500 includes multiple light sources 502, 504, 506, such as multiple display devices (e.g., a desktop computer with multiple monitors), display devices and lamps, multiple display devices and lamps, or similar combinations of light sources. Light sources 502, 504, 506 can be individually controllable. For example, each monitor in an arrangement of multiple monitors can be independently blanked to temporarily reduce light output.

[0066] Glare caused by light sources 502, 504, and 506 can have different characteristics. For example, a monitor directly facing the user 412 can cause glare at the user's eyes, which has a different shape and intensity than glare caused by a monitor angled relative to the user's viewpoint. Furthermore, such a monitor can display different content at different times. For example, during a video conference, a user can have one monitor displaying video of other participants and another monitor displaying a document.

[0067] To train a machine learning model 416 that provides a glare filter, processor 406 can selectively reduce the intensity of multiple light sources 502, 504, 506. That is, processor 406 selects light sources 502, 504, 506 to reduce the intensity during the capture of an image with reduced target brightness. A given target image 430 can be captured using any one or a combination of light sources 502, 504, 506 operating at reduced brightness. Independent modulation of different light sources 502, 504, 506 can provide additional brightness information to machine learning model 416 to increase the accuracy of model 416 in filtering glare. In other examples, each light source 502, 504, 506 can be associated with an independent machine learning model 416 for filtering glare caused by that light source 502, 504, 506.

[0068] In various examples, additional information can be provided to machine learning models to help characterize glare and thereby filter it. Examples of additional information include backlight brightness and light information about the displayed content, such as color and intensity. The light information can be averaged over a region of the display device, over the entire display area, or detailed pixel data can be provided.

[0069] In various examples, the captured images may include visible light, infrared light, or both. Processing infrared images or the infrared component of an image to filter infrared glare can be useful in helping to remove fire glare through downstream processes.

[0070] Given the foregoing, it should be clear that controlling light sources, such as display devices, to temporarily reduce their output can be used to train filters to handle glare that may be caused by the light source. This reduces distraction caused by glare in the captured video and increases the quality of such video. Consequently, video conferences can appear more natural and realistic, especially when users or other objects are susceptible to glare, such as from wearing glasses.

[0071] It should be understood that the features and aspects of the various examples provided above can be combined into other examples that also fall within the scope of this disclosure. Furthermore, the accompanying drawings are not drawn to scale and may have enlarged dimensions and shapes for illustrative purposes.

Claims

1. A non-transitory machine-readable medium comprising instructions for: During a video conference, capture the first image of the scene, including the light emitted by the display device; Change the brightness of the display device; A second image of the scene is captured simultaneously with the change in brightness of the display device; A machine learning model is trained using the first image and the second image to provide a filter to reduce glare generated by light emitted by the display device; as well as The machine learning model is applied to a captured third image of the scene to reduce glare in the third image, which is different from the first and second images.

2. The non-transitory machine-readable medium according to claim 1, wherein, The instruction is used to reduce the brightness of the display device by turning off the backlight of the display device.

3. The non-transitory machine-readable medium according to claim 1, wherein, The instructions are used to periodically reduce the brightness of the display device, capture the second image, and train the machine learning model during a video conference using the display device.

4. The non-transitory machine-readable medium according to claim 3, wherein, The instructions are used to control the frequency of the interval.

5. The non-transitory machine-readable medium according to claim 4, wherein, The instructions are used to control the frequency of the interval based on an error function, wherein a larger error increases the frequency.

6. The non-transitory machine-readable medium according to claim 3, wherein, The instruction is used to trigger a reduction in the brightness of the display device and the capture of the second image based on the content displayed in the video conference.

7. The non-transitory machine-readable medium according to claim 1, wherein, The first image, the second image, and the third image are frames of a video, and the instruction is used to reduce the brightness of the display device for the duration of one frame.

8. An apparatus comprising: light source; Camera; as well as A processor, connected to the light source and the camera, is used for: During a video conference, the camera is controlled to capture a sequence of images; The intensity of the light source is reduced during the capture of the target image of the sequence; A machine learning model is trained using the target image and another image of the sequence to provide a filter to reduce glare, which is generated by light emitted from the light source; as well as The machine learning model is applied to subsequent images in the sequence to reduce glare in the subsequent images.

9. The device of claim 8, further comprising a network interface connected to the processor, wherein: The light source is a display device; The camera is a network camera; and The processor is used to provide video conferencing with display devices, network cameras, and network interfaces.

10. The device according to claim 9, wherein, The processor is used to capture the target image when the video conference is triggered.

11. The device according to claim 8, comprising a plurality of light sources, wherein, The processor is used to selectively reduce the intensity of the plurality of light sources during the capture of the target image.

12. The device according to claim 8, wherein, The machine learning model includes a convolutional neural network.

13. A method comprising: During a video conference, capture the first image of the scene, including light emitted by a light source; Control the light source to output a changed light intensity; A second image is captured from the scene when illuminated by a changing light intensity; A machine learning model is trained using the first image and the second image; as well as The machine learning model is applied to a captured third image of the scene to reduce glare in the third image, which is generated by light emitted from the light source.

14. The method of claim 13, further comprising operating a video conference, wherein, The light source is a display device operated by a user during the video conference, and the first image, the second image, and the third image are captured by the user's camera during the video conference.

15. The method according to claim 14, wherein, Controlling the light source to output the changed light intensity includes blanking the display device.

Images