AUTOMATIC GENERATION OF VIDEO AND DIRECTIONAL AUDIO FROM SPHERICAL CONTENT
1. A method for generating a video with corresponding audio, the method performing by a computing system including one or more processors, the method comprising:
- receiving, by the computing system, video frames having a field of view;
receiving, by the computing system, directional audio captured with the video frames;
determining, by the computing system, a time-varying path of a target within the video frames based on an analysis of content of the video frames or information associated with the video frames;
identifying, by the computing system, sub-frames within the video frames based on the time-varying path of the target, the sub-frames including the target and having a reduced field of view relative to the field of view;
generating, by the computing system, an audio track based on the time-varying path of the target, the audio track including portions of the directional audio corresponding to a direction of the target; and
generating, by the computing system, an output video including the sub-frames and the audio track.
A spherical content capture system captures spherical video and audio content. In one embodiment, captured metadata or video/audio processing is used to identify content relevant to a particular user based on time and location information. The platform can then generate an output video from one or more shared spherical content files relevant to the user. The output video may include a non-spherical reduced field of view such as those commonly associated with conventional camera systems. Particularly, relevant sub-frames having a reduced field of view may be extracted from each frame of spherical video to generate an output video that tracks a particular individual or object of interest. For each sub-frame, a corresponding portion of an audio track is generated that includes a directional audio signal having a directionality based on the selected sub-frame.
- 1. A method for generating a video with corresponding audio, the method performing by a computing system including one or more processors, the method comprising:
receiving, by the computing system, video frames having a field of view; receiving, by the computing system, directional audio captured with the video frames; determining, by the computing system, a time-varying path of a target within the video frames based on an analysis of content of the video frames or information associated with the video frames; identifying, by the computing system, sub-frames within the video frames based on the time-varying path of the target, the sub-frames including the target and having a reduced field of view relative to the field of view; generating, by the computing system, an audio track based on the time-varying path of the target, the audio track including portions of the directional audio corresponding to a direction of the target; and generating, by the computing system, an output video including the sub-frames and the audio track.
- View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
- 9. A non-transitory computer-readable storage medium storing instructions for generating a video with corresponding audio, the instructions when executed by one or more processors causing the one or more processors to perform steps including:
receiving video frames having a field of view; receiving directional audio captured with the video frames; determining a time-varying path of a target within the video frames based on an analysis of content of the video frames or information associated with the video frames; identifying sub-frames within the video frames based on the time-varying path of the target, the sub-frames including the target and having a reduced field of view relative to the field of view; generating an audio track based on the time-varying path of the target, the audio track including portions of the directional audio corresponding to a direction of the target; and generating an output video including the sub-frames and the audio track.
- View Dependent Claims (10, 11, 12, 13, 14)
- 15. A system for generating a video with corresponding audio, the system comprising:
one or more processors; and a non-transitory computer-readable storage medium storing instructions that when executed by the one or more processors causes the one or more processors to perform steps including; receiving video frames having a field of view; receiving directional audio captured with the video frames; generating an output video including the sub-frames and the audio track.
- View Dependent Claims (16, 17, 18, 19, 20)
This disclosure relates to a media content system, and more specifically, to a media content system using spherical video and directional audio.
In a spherical video capture system, a video camera system (which may include multiple video cameras) captures video in a 360 degree field of view along a horizontal axis and 180 degree field of view along the vertical axis, thus capturing the entire environment around the camera system in every direction. Current spherical video systems have not gained widespread use because high resolution, high frame rate video captured by such systems are extremely large and difficult to process and manage.
The disclosed embodiments have other advantages and features which will be more readily apparent from the following detailed description of the invention and the appended claims, when taken in conjunction with the accompanying drawings, in which:
The figures and the following description relate to preferred embodiments by way of illustration only. It should be noted that from the following discussion, alternative embodiments of the structures and methods disclosed herein will be readily recognized as viable alternatives that may be employed without departing from the principles of what is claimed.
Reference will now be made in detail to several embodiments, examples of which are illustrated in the accompanying figures. It is noted that wherever practicable similar or like reference numbers may be used in the figures and may indicate similar or like functionality. The figures depict embodiments of the disclosed system (or method) for purposes of illustration only. One skilled in the art will readily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles described herein.
A spherical content capture system captures spherical video content. A microphone array system captures audio concurrently with the video and produces a plurality of audio channels comprising directional audio signals corresponding to different directions. A spherical content sharing platform enables users to share the captured spherical content and enables users to access spherical content shared by other users. By definition, a spherical camera will capture everything in the surrounding environment (e.g., 360 degrees in the horizontal plane and 180 degrees in the vertical plane). While only a small portion of the captured content may be relevant to operator of the camera, the remainder of the captured content may be relevant to a community of other users. For example, any individuals that were in the vicinity of a spherical camera capturing spherical video content are likely to appear somewhere in the captured content, and may therefore be interested in the content. Thus, any captured spherical content may be meaningful to a number of different individuals and a community of users may benefit from sharing of spherical video content. As one example, a group of people each record their actions on a spherical camera and allow shared access to the captured content. Each individual in the group is then capable of extracting relevant and meaningful content from a shared capture, different portions of which will be relevant to different members of the group or others outside of the group.
In one embodiment, metadata included in the spherical content is used to identify content relevant to a particular user (e.g., based on time and location information). Alternatively, the content relevant to a particular user may be determined by processing the spherical video itself (e.g., by performing content recognition) or its corresponding audio channels. The platform can then generate an output video from one or more shared spherical content files relevant to the user. The output video may include a non-spherical reduced field of view such as those commonly associated with conventional camera systems (e.g., a 120 degree by 67 degree field of view). For example, a relevant sub-frame having a reduced field of view may be extracted from each frame of spherical video to generate the output video. For example, sub-frames may be selected to generate an output video that track a particular individual, object, scene, or activity of interest. The output video thus reduces the captured spherical content to a standard field of view video having the content of interest while eliminating extraneous data outside the targeted field of view.
As with the output video, the output audio may have a smaller audible field or may be directionally weighted such that audio from a particular direction or range of directions is more prominent. As the viewing window changes in a captured spherical scene, the audio changes accordingly. Thus, an audio track may be generated to correspond to the output video such that for each sub-frame of the of the output video, the directionality associated with the audio track corresponds to the location of the selected sub-frame. For example, if the location of the extracted sub-frames are selected to simulate a left-to-right pan across a scene, the audio track will be generated in a manner that simulates a microphone panning from left to right (i.e., weighting audio from the left side of the scene more heavily at the beginning of the pan and weighting audio from the right side of the scene more heavily at the end of the pan). Furthermore, if the video is zoomed to reduce the field of view, the audio may similarly be adjusted to be more directionally focused in the zoomed direction. Thus, for example, sound originating from a subject shown in the output video may be heard more prominently than sound originating from a subject outside the field of view. In one embodiment, audio from different directions is weighted in order to create a realistic audio experience. For example, audio from directions other than where the viewer is focused may be present in the recreated audio, but the various channels may be weighted such that the audio in the viewing direction is most prominent. Thus, the output video simulates both visually and audibly a video captured with a standard field of view camera that is manually pointed to track an individual, object, or scene of interest. As will be apparent, many different output videos can be generated from the same set of shared spherical video content.
In a particular embodiment, a method, video server, and non-transitory computer-readable storage medium generates a video with corresponding audio. A spherical video is received. A plurality of audio channels is also received representing audio captured concurrently with the spherical video, where each of the plurality of audio channels comprising directional audio corresponding to a different direction. Different spatial regions of the spherical video are mapped to different audio channels. For each of a plurality of frames of the spherical video, a sub-frame is extracted having a reduced field of view relative to a field of view of the spherical video. For each of the sub-frames, a processor determines one or more of the spatial regions overlapping the sub-frame and generates a portion of an audio stream from one or more different audio channels corresponding to the one or more spatial regions overlapping the sub-frame. The generated video and the audio stream is then outputted.
Additional embodiments are described in further detail below.
Generation of Output Video from Spherical Content
In one embodiment, a community content sharing platform enables individuals to subscribe to a community of users. The subscribers are provided access to video captured by not only themselves but also the wider group. The community content sharing platform may effectively be a public open-source resource for everyone to find and use meaningful content of themselves from a plurality of different spherical camera sources.
The camera 230 can include a camera body one or more a camera lenses, various indicators on the camera body (such as LEDs, displays, and the like), various input mechanisms (such as buttons, switches, and touch-screen mechanisms), and electronics (e.g., imaging electronics, power electronics, metadata sensors, etc.) internal to the camera body for capturing images via the one or more lenses and/or performing other functions. One or more cameras 230 is capable of capturing spherical or substantially spherical content and capturing audio via a multi-directional microphone array. As used herein, spherical content may include still images or video having spherical or substantially spherical field of view. For example, in one embodiment, the camera 230 captures video having a 360 degree field of view in the horizontal plane and a 180 degree field of view in the vertical plane. Alternatively, the camera 230 may capture substantially spherical video having less than 360 degrees in the horizontal direction and less than 180 degrees in the vertical direction (e.g., within 10% of the field of view associated with fully spherical content).
As described in greater detail in conjunction with
The video server 240 receives and stores videos captured by the camera 230 and allows users to access shared videos at a later time. In one embodiment, the video server 240 provides the user with an interface, such as a web page or native application installed on the client device 235, to interact with and/or edit the stored videos and to automatically generate output videos relevant to a particular user (or a particular set of metadata) from one or more stored spherical videos. The output videos may have a reduced field of view relative to the original spherical videos. For example, an output video may have a field of view consistent with that of a conventional non-spherical camera such as, for example, a 120 degree by 67 degree field of view. To generate the output video, the video server 240 extracts a sequence of relevant sub-frames having the reduced field of view from frames of one or more spherical videos and processes audio signals captured from the microphone array to generate an audio track having directionality that varies based on the selected sub-frames. For example, sub-frames may be selected from one or more spherical videos to generate an output video that tracks a path of a particular individual or object. In one embodiment, the video server 240 can automatically identify sub-frames by finding identifying spherical video that was captured near a particular location and time where a user was present (or other time and location of interest). Because spherical content is captured in all directions, the spherical video captured at the particular time and location where the user was present is highly likely to include sub-frames depicting the user (or scene of interest). In another embodiment, a time-varying path (e.g., a sequence of time-stamped locations) of a target (e.g., a person, object, or other scene of interest) can be used to automatically find spherical video having time and location metadata closely matching the path. Furthermore, by correlating the relative location of the camera 230 with a location at each time point in the path of interest, the video server 240 can automatically determine a direction between the camera 230 and the target and thereby automatically select the appropriate sub-frames depicting the target. Furthermore, an audio track can be generated having a directionality that varies over time based on the determined direction between the camera 230 and the target so as to more heavily weight audio coming from the direction of the target. Because the original spherical video captures video in all directions, many different output videos can be generated from the same set of shared spherical video content.
In an embodiment, the video server 240 generates the output video based on an input metadata that includes a target path represented by a sequence of time-stamped target locations. For example, the person depicted in
As one example use case scenario, a skier at a ski resort may use an application on his mobile phone to track his movement through the ski resort. One or more other users capture spherical video content one the same day at the same ski resort and share the spherical content on the video server, some of which will depict the skier. By correlating the time/location metadata tracking the skier'"'"'s movement with time/location metadata associated with the spherical video files, the video server can automatically locate a sequence of sub-frames from one or more of the spherical videos that depict the skier and follow his path through the resort. Further still, other skiers can input a different set of time/location metadata and obtain their own customized videos from the same set of captured spherical content. If multiple skiers record and share spherical content, the volume of relevant video for any individual skier is multiplied. Thus, as the size of the sharing community increases, the relevance of the spherical content to any giving user increases rapidly.
In other embodiments, the video server 240 can automatically identify sub-frames of interest based on the spherical video content itself or its associated audio track. For example, facial recognition, object recognition, motion tracking, or other content recognition or identification techniques may be applied to the spherical video to identify sub-frames of interest. Alternatively, or in addition, a microphone array may be used to determine directionality associated with a received audio signal, and the sub-frames of interest may be chosen based on the direction between the camera and the audio source. These embodiments beneficially can be performed without any location tracking of the target of interest. Further still, any of the above techniques may be used in combination to automatically determine which sub-frames to select for generating an output video.
A user can interact with interfaces provided by the video server 240 via the client device 235. The client device 235 is any computing device capable of receiving user inputs as well as transmitting and/or receiving data via the network 220. In one embodiment, the client device 235 is a conventional computer system, such as a desktop or a laptop computer. Alternatively, the client device 235 may be a device having computer functionality, such as a personal digital assistant (PDA), a mobile telephone, a smartphone or another suitable device. The user can use the client device 235 to view and interact with or edit videos stored on the video server 240. For example, the user can view web pages including video summaries for a set of videos captured by the camera 230 via a web browser on the client device 235.
One or more input devices associated with the client device 235 receive input from the user. For example, the client device 235 can include a touch-sensitive display, a keyboard, a trackpad, a mouse, a voice recognition system, and the like. In some embodiments, the client device 235 can access video data and/or metadata from the camera 230 or one or more metadata sources 210, and can transfer the accessed metadata to the video server 240. For example, the client device may retrieve videos and metadata associated with the videos from the camera via a universal serial bus (USB) cable coupling the camera 230 and the client device 235. The client device 235 can then upload the retrieved videos and metadata to the video server 240. In one embodiment, the client device 235 interacts with the video server 240 through an application programming interface (API) running on a native operating system of the client device 235, such as IOS® or ANDROID™. While
The video server 240 communicates with the client device 235, the metadata sources 210, and the camera 230 via the network 220, which may include any combination of local area and/or wide area networks, using both wired and/or wireless communication systems. In one embodiment, the network 220 uses standard communications technologies and/or protocols. In some embodiments, all or some of the communication links of the network 220 may be encrypted using any suitable technique or techniques. It should be noted that in some embodiments, the video server 240 is located within the camera 230 itself.
Various components of the environment 200 of
An input/output (I/O) interface 360 transmits and receives data from various external devices. For example, the I/O interface 360 may facilitate the receiving or transmitting video or audio information through an I/O port. Examples of I/O ports or interfaces include USB ports, HDMI ports, Ethernet ports, audioports, and the like. Furthermore, embodiments of the I/O interface 360 may include wireless ports that can accommodate wireless connections. Examples of wireless ports include Bluetooth, Wireless USB, Near Field Communication (NFC), and the like. The I/O interface 360 may also include an interface to synchronize the camera 230 with other cameras or with other external devices, such as a remote control, a second camera 230, a smartphone, a client device 335, or a video server 340.
A control/display subsystem 370 includes various control a display components associated with operation of the camera 230 including, for example, LED lights, a display, buttons, microphones, speakers, and the like.
The audio subsystem 350 includes an array of microphones (e.g., at least two microphones) and one or more audio processors to capture and process audio data correlated with video capture. The microphones are positioned in a manner that enables audio to be captured from various directions. A plurality of audio channels are generated each corresponding to a different direction. In one embodiment, each audio channel simply corresponds to audio captured by one the microphones. For example, in a six microphone array having microphones facing in the up, down, left, right, front, and back directions, six audio channels are generated corresponding to each of these directions. In alternative embodiments, audio beamforming techniques may be used to generate directional audio channels from weighted combinations of different microphones so that, for example, six audio channels corresponding to the up, down, left, right, front, and back directions can be generated from fewer than six microphones. Each audio channel has an associated directionality in that audio originating from a source in the associated direction is more heavily weighted than audio in originating from sources from other directions. For example, when an individual is speaking, an audio channel in the direction of the speaker would include the speaker at relatively high volume relative to background noise, while an audio channel in a direction opposite the speaker may include the speaker at a reduced or inaudible volume.
Sensors 340 capture various metadata concurrently with, or separately from, video capture. For example, the sensors 340 may capture time-stamped location information based on a global positioning system (GPS) sensor, and/or an altimeter. Other sensors 340 may be used to detect and capture orientation of the camera 230 including, for example, an orientation sensor, an accelerometer, a gyroscope, or a magnetometer. Sensor data captured from the various sensors 340 may be processed to generate other types of metadata. For example, sensor data from the accelerometer may be used to generate motion metadata, comprising velocity and/or acceleration vectors representative of motion of the camera 230. Furthermore, sensor data from the may be used to generate orientation metadata describing the orientation of the camera 230. Sensor data from the GPS sensor provides GPS coordinates identifying the location of the camera 230, and the altimeter measures the altitude of the camera 230. In one embodiment, the sensors 340 are rigidly coupled to the camera 230 such that any motion, orientation or change in location experienced by the camera 230 is also experienced by the sensors 340. The sensors 340 furthermore may associates a time stamp representing when the data was captured by each sensor. In one embodiment, the sensors 340 automatically begin collecting sensor metadata when the camera 230 begins recording a video.
In alternative embodiments, one or more components of the camera cores 310 may be shared between different camera cores 310. For example, in one embodiment, the camera cores 310 may share one or more image processors 316. Furthermore, in alternative embodiments, the camera cores 310 may have additional separate components such as, for example, dedicated system memory 330 or system controllers 320. In yet other embodiments, the camera 230 may have more than two camera cores 310 or a single camera core with a 360° lens or a single hyper-hemi (super fish-eye) lens.
In one embodiment, the camera 230 comprises a twin hyper-hemispherical lens system that capture two image hemispheres with synchronized image sensors which combine to form a contiguous spherical image. The image hemispheres may be combined based on, for example, a back-to-back configuration, a side-by-side configuration, a folded symmetrical configuration or a folded asymmetrical configuration. Each of the two streams generated by camera cores 310 may be separately encoded and then aggregated in post processing to form the spherical video. For example, each of the two streams may be encoded at 2880×2880 pixels at 30 frames per second and combined to generate a 5760×2880 spherical video at 30 frames per second. Other resolutions and frame rates may also be used.
In an embodiment the spherical content is captured at a high enough resolution to guarantee the desired output from the relevant sub-frame will be of sufficient resolution. For example, if a horizontal field of view of 120° at an output resolution of 1920×1080 pixels is desired in the final output video, the original spherical capture may include a horizontal 360° resolution of at least 5760 pixels (3×1920).
In one embodiment, a 5.7K spherical file format provides 16 megapixel resolution. This provides a resolution of approximately one pixel per inch at a distance of 23 meters (76 feet) from the camera 230. In this embodiment, spherical video is captured as a 5760 pixels by 2880 pixels with a 360 degree horizontal field of view and a 180 degree vertical field of view. In one embodiment, the image sensor may capture 6 k×3 k image to provide six degrees of overlap and 4 degrees of out-of-field image to avoid worst modulation transfer function (MTF) region from the lens. From the spherical image frames, a 1920×1080 sub-frame may be extracted that provides a 120 degree by 67.5 degree field of view. As described above, the location of the sub-frame may be selected to capture sub-frames of interest to a given user. In one embodiment, each of two image sensors capture 3 k×3 k images which are encoded as 2880×2880 images, which are combined to create the 5760×2880 spherical image.
In another embodiment, a 720p file format is used. Here, spherical video is represented as 4000 pixels by 2000 pixels with a 360 degree horizontal field of view and a 180 degree vertical field of view. In one embodiment, the 4 k×2 k image may be based on a 4000 pixels×2250 pixels image captured by the image sensor to provide some overlap in the vertical direction. From the spherical image frames, a 720×1280 sub-frame may be extracted from each frame that provides a 115 degree by 65 degree field of view.
In one embodiment, the camera 230 includes a computational image processing chip that aggregates the two data streams into one encoding internally to the camera 230. The camera 230 can then directly output the spherical content or a downscaled version of it. Furthermore, in this embodiment, the camera 230 may directly output sub-frames of the captured spherical content having a reduced field of view based on user control inputs specifying the desired sub-frame locations.
In an embodiment, the video server 240 enables users to create and manage individual user accounts. User account information is stored in the user storage 505. A user account may include information provided by the user (such as biographic information, geographic information, and the like) and may also include additional information inferred by the video server 240 (such as information associated with a user'"'"'s historical use of a camera and interactions with the video server 240). Examples of user information include a username, contact information, a user'"'"'s hometown or geographic region, other location information associated with the user, other users linked to the user as “friends,” and the like. The user storage 505 may include data describing interactions between a user and videos captured by the user. For example, a user account can include a unique identifier associating videos uploaded by the user with the user'"'"'s user account. Furthermore, the user account can include data linking the user to other videos associated with the user even if the user did not necessarily provide those videos. For example, the user account may link the user to videos having location metadata matching the user'"'"'s location metadata, thus indicating that the video was captured at a time and place where the user was present and the user is therefore highly likely to be depicted somewhere in the video.
The video storage 510 stores videos captured and uploaded by users of the video server 240. The video server 240 may access videos captured using the camera 230 and store the videos in the video storage 510. In one example, the video server 240 may provide the user with an interface executing on the client device 235 that the user may use to upload videos to the video storage 515. In one embodiment, the video server 240 indexes videos retrieved from the camera 230 or the client device 235, and stores information associated with the indexed videos in the video store. For example, the video server 240 provides the user with an interface to select one or more index filters used to index videos. Examples of index filters include but are not limited to: the time and location that the video was captured, the type of equipment used by the user (e.g., ski equipment, mountain bike equipment, etc.), the type of activity being performed by the user while the video was captured (e.g., snowboarding, mountain biking, etc.), or the type of camera 230 used to capture the content.
In some embodiments, the video server 240 generates a unique identifier for each video stored in the video storage 510 which may be stored as metadata associated with the video in the metadata storage 525. In some embodiments, the generated identifier for a particular video is unique to a particular user. For example, each user can be associated with a first unique identifier (such as a 10-digit alphanumeric string), and each video captured by a user is associated with a second unique identifier made up of the first unique identifier associated with the user concatenated with a video identifier (such as an 8-digit alphanumeric string unique to the user). Thus, each video identifier is unique among all videos stored at the video storage 510, and can be used to identify the user that captured the video.
In some embodiment, in addition to being associated with a particular user, a video may be associated with a particular community. For example, the video provider may choose to make the video private, make the video available with the entire public, or make the video available to one or more limited specified community such as, for example, the user'"'"'s friends, co-workers, members in a particular geographic region, etc.
The metadata storage 525 stores metadata associated with videos stored by the video storage 510 and with users stored in the user storage 505. Particularly, for each video, the metadata storage 525 stores metadata including time-stamped location information associated with each frame of the video to indicate the location of the camera 230 at any particular moment during capture of the spherical content. Additionally, the metadata storage 525 may store other types of sensor data captured by the camera 230 in association with a video including, for example, gyroscope data indicating motion and/or orientation of the device. In some embodiments, metadata corresponding to a video is stored within a video file itself, and not in a separate storage module. The metadata storage 525 may also store time-stamped location information associated with a particular user so as to represent a user'"'"'s physical path during a particular time interval. This data may be obtained from a camera held by the user, a mobile phone application that tracks the user'"'"'s path, or another metadata source.
The web server 530 provides a communicative interface between the video server 240 and other entities of the environment of
A video pre-processing module 560 pre-processes and indexes uploaded videos. For example, in one embodiment, uploaded videos are automatically processed by the video pre-processing module 560 to conform the videos to a particular file format, resolution, etc. Furthermore, in one embodiment, the video pre-processing module 560 may automatically parse the metadata associated with videos upon being uploaded to determine the temporal and spatial path of the camera. The video pre-processing module 560 may furthermore automatically correlate the camera path to path data associated with the users in order to identify users that were in the vicinity of the camera when it was capturing the spherical content and are therefore likely to appear somewhere in the spherical video. In an embodiment, a user may be automatically alerted when new spherical video is uploaded that has metadata closely matching the user'"'"'s metadata, thus indicating the video was taken in the vicinity of the user. Additionally, the video pre-processing module 560 may automatically perform audio and/or video processing algorithms to identify sub-frames of interest independently of known target locations. For example, the video pre-processing module 560 may perform facial recognition and facial tracking to identify an individual in a spherical video and tag that person in the video. Additionally, content recognition may be performed to identify particular objects in the video. In another embodiment, motion detection may be performed to identify regions of the video having certain motion parameters that may represent an activity of interest. In yet another embodiment, gesture recognition may be performed to identify regions of the video having an individual performing a particular gesture. In yet another embodiment, audio processing and/or speech recognition may be performed based on a microphone array input to identify locations of one or more audio sources in the spherical video.
The video generation module 540 automatically generates output videos relevant to a user or to a particular set of inputs. For example, the video generation module 540 may generate an output video including content that tracks a sequence of locations representing a physical path over a particular time interval and produces a corresponding audio track that varies the directionality of the captured audio based on the tracked sequence of locations. Alternatively, the video generation module 540 may generate an output video including content that tracks a particular face or object identified in the video, tracks an area of motion having particular motion characteristics, tracks an identified audio source, etc. The output videos have a reduced field of view (e.g., a standard non-spherical field of view) and represent relevant sub-frames to provide a video of interest. For example, the video may track a particular path of an individual, object, or other target so that each sub-frame depicts the target as the target moves through a given scene. In one embodiment, the video generation module 540 may operate in response to a user querying the video server 240 with particular input criteria. In another embodiment, the video generation module 540 may automatically generate videos relevant to users of the community based on metadata or profile information associated with user and automatically provide the videos to the user when it is identified as being relevant to the user (e.g., via their web portal, via email, via text message, or other means).
In an embodiment, the video server 240 enables the user to select from predefined video generation templates. For example, the user can request that the video server 240 generate a video based on location tracking, based on facial recognition, gesture recognition, audio tracking, motion detection, or other technique. Various parameters used by the video server 240 to select relevant frames such as thresholds governing proximity distance and clip duration can be adjusted or pre-set.
In an embodiment, content manipulation is performed on the video server 240 with edits and playback using only the original source content. In this embodiment, when generating an output video, the video server 140 saves an edit map indicating, for each frame of the output video, the original spherical video file from which the sub-frame was extracted and the location of the sub-frame. The edit map may furthermore store any processing edits performed on the video such as, for example, image warping, image stabilization, output window orientation, image stitching changes in frame rate or formatting, audio mixing, effects, etc. In this embodiment, no copying, storing, or encoding of a separate output video sequences is necessary. This beneficially minimizes the amount of data handled by the server. When users views a previously saved output video, the server 240 re-generates the output video based on the saved edit map by retrieving the relevant sub-frames from the original source content and regenerates a corresponding audio track with directionality based on the selected sub-frames. Alternatively, the user may select to download a copy of the output video, for storing in the user'"'"'s local storage.
In an embodiment, the user interface also provides an interactive viewer that enables the user to pan around within the spherical content being viewed. This will allow the user to search for significant moments to incorporate into the output video and manually edit the automatically generated video.
In one embodiment, the user interface enables various editing effects to be added to a generated output video. For example, the video editing interface may enable effects such as, cut-away effects, panning, tilting, rotations, reverse angles, image stabilization, zooming, object tracking,
In one embodiment, spherical content may also be processed to improve quality. For example, in one embodiment, dynamic stabilization is applied to stabilize in the horizontal, vertical, and rotational directions. Because the content is spherical, stabilization can be performed with no loss of image resolution. Stabilization can be performed using various techniques such as object tracking, vector map analysis, on-board gyro data, etc. For example, an in-view camera body can be used as a physical or optical reference for stabilization. Spherical content may also be processed to reduce rolling shutter artifacts. This may be performed using on-board gyro motion data or image analysis data. This processing is also lossless (i.e., no pixels are pushed out of the frame.). In this technique, horizontal pixel lines are rotated to re-align an image with the true vertical orientation. The technique works for rotational camera motion within an environment (e.g., when the camera is spinning).
In one embodiment, to encourage users to share content, the platform may reward the user with credits when his/her content is accessed/used by other members of a group or community. Furthermore, a user may spend credits to access other content streams on the community platform. In this way, users are incentivized to carry a camera and to capture compelling content. If socially important spherical content is made available by a particular user, the user could generate an income-stream as people access that content and post their own edits.
In another embodiment, an output video may be generated from two or more different spherical videos. For example, two or more different portions (which may be overlapping) of the user metadata may have matches against two different spherical videos. A first portion of the output video can then be generated by extracting relevant sub-frames from the corresponding first spherical video and a second portion of the output video can be generated by extracting relevant sub-frames from the corresponding second spherical video. If the matching portions overlap, the video server 240 may automatically determine when to cut from the first spherical video to the second spherical video. For example, the video server 240 may to determine to transition between the videos in a manner that maintains the shortest distance between the camera and the target. As the number of shared spherical videos within a given vicinity increases, users may be able to generate increasingly longer output videos tracking a given target path through an environment by combining relevant sub-frames from different spherical videos into a single combined output video.
Other techniques for identifying relevant sub-frames do not necessarily depend on location data associated with the target and instead identify sub-frames relevant to a particular target based on the spherical media content (e.g., visual and/or audio content) itself.
In another embodiment, audio detected using directional microphones can be coupled with faces or other recognized visual elements that are potential audio sources. For example, in one embodiment, if a directionality of an audio signal is determined to sufficiently correspond to a location of a face (detected via a face detection algorithm) or other sound source, the audio in the output video can be weighted based on the determined scene motion in the view window such that the audio experience is associated with the visual elements, i.e., the directionality of the audio signal tracks the direction of motion of the discovered sound source. Furthermore, the directional audio signal may be used to further improve a scene motion analysis algorithm (such as face or object tracking) by weighting the prediction location of the face or object more heavily when it is well-correlated to the change in audio directionality.
Furthermore, in one embodiment, speech recognition may be used to differentiate a sound of interest from background noise. For example, a user may speak a command such as “tag me” or state the user'"'"'s name to indicate that the user'"'"'s location in the video. In other embodiments, a location of a target feature may be manually identified. In yet further embodiments, two or more of the techniques described above can be combined to identify a target feature of interest. For example, in one embodiment, different regions of the video may be scored based on a number of weighted metrics and a region corresponding to a target feature is chosen based on the weighted score.
Based on the identified target, a range of spherical frames is identified 806 that includes the target feature. For example, in one embodiment the range of spherical frames meets at least some minimum time duration to ensure that a meaningful length of video can be extracted. Sub-frames of the spherical video around the spatial location of the target feature are then determined 808 within the identified range. In one embodiment, sub-frame selection is performed in order to generally provide smooth transitions between frames of the output video (i.e., consecutive sub-frames in the output video generally have spatial regions within some small distance threshold of each other) to provide an output video that closely resembles the output of a camera that pans across a scene without discontinuous jumps. An audio track is generated 810 having directionality corresponding to the relative orientation between the camera and the target feature of interest. An output video is then generated 812 from the relevant sub-frames and audio track in order to produce an output video having a reduced field of view and corresponding directional audio tracking the object or region of interest.
In another embodiment, the video server 240 can generate an output video from a statically located sub-frame that corresponds to a center region (or other predefined region) of one the image captured by one of the lenses. If the user knows which lens is considered the “forward-looking” (e.g., based on a physical feature of the camera body), the user can “point” the spherical camera so that the forward-looking lens is oriented in the direction of interest. In other words, the user points the spherical camera in the same way that a user would operate a standard camera with a standard (non-spherical) field-of-view. In this way, a video can be generated that would be equivalent to what would have been captured by a camera having a standard (non-spherical) field-of-view. However, because the camera captures in all directions, the remaining content outside the center sub-frames may be of interest in generating other secondary videos using the techniques described above.
As described above, different portions of a given spherical video may be relevant to a large number of different users of the sharing community. In one embodiment, rather than the video server 240 storing individual output videos generated for each of its users, the video server can instead store an edit map specifying how the desired output video can be regenerated from the original raw spherical video and/or how the audio track can be reconstructed from the available audio channels (e.g., storing pointers to the audio channels and associated weights at different sample times). Then, the output video can be generated on request (e.g., in real-time) from the edit map when a user requests viewing. For example, the output video can be streamed to the user or the user can download the output video to the user'"'"'s own local storage. An advantage of this approach is that individual output videos for specific users need not be stored by the video server 240, thus reducing its storage requirements. This storage savings may be significant because it is expected that a large number of personalized output videos may be generated from a relatively small number of shared spherical videos.
For audio, the edit map may specify how to reconstruct the output audio track from the stored audio channels by indicating the weights to apply to each audio channel at different time instances. Alternatively, the video server 240 may automatically determine the weights when the video is requested based on the selected frames.
In one embodiment, the raw spherical video is stored by the video server 240 as two or more unstitched portions. For example, if the spherical video is captured with a camera having two hemispherical lenses as described above, the two hemispherical videos are stored by the video server 240 as separate raw files in an unstitched form. Then, the videos may be stitched together when generating the output video only if a selected sub-frames includes image data from more than one unstitched portion of the spherical content. Beneficially, if a given sub-frame is entirely within one hemisphere, no stitching need be performed for that sub-frame. Furthermore, where stitching is performed, the size of the edge for stitching is limited by the dimensions of the sub-frame and processing is substantially reduced compared to performing stitching on the entire spherical video.
As described above, the edit map may be generated automatically based on metadata or based on audio/visual processing to identify relevant target features. Alternatively, edit maps may be generated manually using a spherical video viewer and editing tools available through the video server 240. For example, in one embodiment, the video server 240 displays a selected spherical video in a spherical video viewer and receives user inputs indicating selections of the desired sub-frames at each frame of the selected spherical video. The edit map can then be specified based on the user selections.
In one embodiment, previously generated output videos (either automatically or manually generated) may be available for manual editing by users in order to further refine the output videos. For example, users may edit the videos to adjust the sub-frame location at each frame, change cut locations between videos when multiple overlapping matching portions exist, or apply other video processing effects.
As illustrated in
To address the problem above, in one embodiment, an audio map may be generated that links the different audio channels to different spatial regions (or “audio zones”) of the spherical video. As shown in
It is also apparent from
In one embodiment, the audio output always includes contributions for each of the audio channels. In this embodiment, the channels may be weighted so that some directions are more prominent, but none of the channels reduce to zero.
In one embodiment, a stereo or surround audio output can be synthesized using the mapping technique described above.
Throughout this specification, some embodiments have used the expression “coupled” along with its derivatives. The term “coupled” as used herein is not necessarily limited to two or more elements being in direct physical or electrical contact. Rather, the term “coupled” may also encompass two or more elements are not in direct contact with each other, but yet still co-operate or interact with each other, or are structured to provide a thermal conduction path between the elements.
Likewise, as used herein, the terms “comprises,” “comprising,” “includes,” “including,” “has,” “having” or any other variation thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, article, or apparatus that comprises a list of elements is not necessarily limited to only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
In addition, use of the “a” or “an” are employed to describe elements and components of the embodiments herein. This is done merely for convenience and to give a general sense of the invention. This description should be read to include one or at least one and the singular also includes the plural unless it is obvious that it is meant otherwise.
Finally, as used herein any reference to “one embodiment” or “an embodiment” means that a particular element, feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment.
Upon reading this disclosure, those of skill in the art will appreciate still additional alternative structural and functional designs for the described embodiments as disclosed from the principles herein. Thus, while particular embodiments and applications have been illustrated and described, it is to be understood that the disclosed embodiments are not limited to the precise construction and components disclosed herein. Various modifications, changes and variations, which will be apparent to those skilled in the art, may be made in the arrangement, operation and details of the method and apparatus disclosed herein without departing from the scope defined in the appended claims.