henrike@webrtc.org | 28e2075 | 2013-07-10 00:45:36 +0000 | [diff] [blame] | 1 | // libjingle |
| 2 | // Copyright 2010 Google Inc. |
| 3 | // |
| 4 | // Redistribution and use in source and binary forms, with or without |
| 5 | // modification, are permitted provided that the following conditions are met: |
| 6 | // |
| 7 | // 1. Redistributions of source code must retain the above copyright notice, |
| 8 | // this list of conditions and the following disclaimer. |
| 9 | // 2. Redistributions in binary form must reproduce the above copyright notice, |
| 10 | // this list of conditions and the following disclaimer in the documentation |
| 11 | // and/or other materials provided with the distribution. |
| 12 | // 3. The name of the author may not be used to endorse or promote products |
| 13 | // derived from this software without specific prior written permission. |
| 14 | // |
| 15 | // THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED |
| 16 | // WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF |
| 17 | // MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO |
| 18 | // EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 19 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| 20 | // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; |
| 21 | // OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, |
| 22 | // WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR |
| 23 | // OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF |
| 24 | // ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 25 | // |
| 26 | // Declaration of abstract class VideoCapturer |
| 27 | |
| 28 | #ifndef TALK_MEDIA_BASE_VIDEOCAPTURER_H_ |
| 29 | #define TALK_MEDIA_BASE_VIDEOCAPTURER_H_ |
| 30 | |
| 31 | #include <string> |
| 32 | #include <vector> |
| 33 | |
| 34 | #include "talk/base/basictypes.h" |
| 35 | #include "talk/base/criticalsection.h" |
| 36 | #include "talk/base/messagehandler.h" |
| 37 | #include "talk/base/scoped_ptr.h" |
| 38 | #include "talk/base/sigslot.h" |
| 39 | #include "talk/base/thread.h" |
| 40 | #include "talk/media/base/videocommon.h" |
| 41 | #include "talk/media/devices/devicemanager.h" |
| 42 | |
| 43 | |
| 44 | namespace cricket { |
| 45 | |
| 46 | class VideoProcessor; |
| 47 | |
| 48 | // Current state of the capturer. |
| 49 | // TODO(hellner): CS_NO_DEVICE is an error code not a capture state. Separate |
| 50 | // error codes and states. |
| 51 | enum CaptureState { |
| 52 | CS_STOPPED, // The capturer has been stopped or hasn't started yet. |
| 53 | CS_STARTING, // The capturer is in the process of starting. Note, it may |
| 54 | // still fail to start. |
| 55 | CS_RUNNING, // The capturer has been started successfully and is now |
| 56 | // capturing. |
| 57 | CS_PAUSED, // The capturer has been paused. |
| 58 | CS_FAILED, // The capturer failed to start. |
| 59 | CS_NO_DEVICE, // The capturer has no device and consequently failed to start. |
| 60 | }; |
| 61 | |
| 62 | class VideoFrame; |
| 63 | |
| 64 | struct CapturedFrame { |
| 65 | static const uint32 kFrameHeaderSize = 40; // Size from width to data_size. |
| 66 | static const uint32 kUnknownDataSize = 0xFFFFFFFF; |
| 67 | |
| 68 | CapturedFrame(); |
| 69 | |
| 70 | // Get the number of bytes of the frame data. If data_size is known, return |
| 71 | // it directly. Otherwise, calculate the size based on width, height, and |
| 72 | // fourcc. Return true if succeeded. |
| 73 | bool GetDataSize(uint32* size) const; |
| 74 | |
| 75 | // The width and height of the captured frame could be different from those |
| 76 | // of VideoFormat. Once the first frame is captured, the width, height, |
| 77 | // fourcc, pixel_width, and pixel_height should keep the same over frames. |
| 78 | int width; // in number of pixels |
| 79 | int height; // in number of pixels |
| 80 | uint32 fourcc; // compression |
| 81 | uint32 pixel_width; // width of a pixel, default is 1 |
| 82 | uint32 pixel_height; // height of a pixel, default is 1 |
| 83 | int64 elapsed_time; // elapsed time since the creation of the frame |
| 84 | // source (that is, the camera), in nanoseconds. |
| 85 | int64 time_stamp; // timestamp of when the frame was captured, in unix |
| 86 | // time with nanosecond units. |
| 87 | uint32 data_size; // number of bytes of the frame data |
| 88 | int rotation; // rotation in degrees of the frame (0, 90, 180, 270) |
| 89 | void* data; // pointer to the frame data. This object allocates the |
| 90 | // memory or points to an existing memory. |
| 91 | |
| 92 | private: |
| 93 | DISALLOW_COPY_AND_ASSIGN(CapturedFrame); |
| 94 | }; |
| 95 | |
| 96 | // VideoCapturer is an abstract class that defines the interfaces for video |
| 97 | // capturing. The subclasses implement the video capturer for various types of |
| 98 | // capturers and various platforms. |
| 99 | // |
| 100 | // The captured frames may need to be adapted (for example, cropping). Adaptors |
| 101 | // can be registered to the capturer or applied externally to the capturer. |
| 102 | // If the adaptor is needed, it acts as the downstream of VideoCapturer, adapts |
| 103 | // the captured frames, and delivers the adapted frames to other components |
| 104 | // such as the encoder. Effects can also be registered to the capturer or |
| 105 | // applied externally. |
| 106 | // |
| 107 | // Programming model: |
| 108 | // Create an object of a subclass of VideoCapturer |
| 109 | // Initialize |
| 110 | // SignalStateChange.connect() |
| 111 | // SignalFrameCaptured.connect() |
| 112 | // Find the capture format for Start() by either calling GetSupportedFormats() |
| 113 | // and selecting one of the supported or calling GetBestCaptureFormat(). |
| 114 | // Start() |
| 115 | // GetCaptureFormat() optionally |
| 116 | // Stop() |
| 117 | // |
| 118 | // Assumption: |
| 119 | // The Start() and Stop() methods are called by a single thread (E.g., the |
| 120 | // media engine thread). Hence, the VideoCapture subclasses dont need to be |
| 121 | // thread safe. |
| 122 | // |
| 123 | class VideoCapturer |
| 124 | : public sigslot::has_slots<>, |
| 125 | public talk_base::MessageHandler { |
| 126 | public: |
| 127 | typedef std::vector<VideoProcessor*> VideoProcessors; |
| 128 | |
| 129 | // All signals are marshalled to |thread| or the creating thread if |
| 130 | // none is provided. |
| 131 | VideoCapturer(); |
| 132 | explicit VideoCapturer(talk_base::Thread* thread); |
| 133 | virtual ~VideoCapturer() {} |
| 134 | |
| 135 | // Gets the id of the underlying device, which is available after the capturer |
| 136 | // is initialized. Can be used to determine if two capturers reference the |
| 137 | // same device. |
| 138 | const std::string& GetId() const { return id_; } |
| 139 | |
| 140 | // Get the capture formats supported by the video capturer. The supported |
| 141 | // formats are non empty after the device has been opened successfully. |
| 142 | const std::vector<VideoFormat>* GetSupportedFormats() const; |
| 143 | |
| 144 | // Get the best capture format for the desired format. The best format is the |
| 145 | // same as one of the supported formats except that the frame interval may be |
| 146 | // different. If the application asks for 16x9 and the camera does not support |
| 147 | // 16x9 HD or the application asks for 16x10, we find the closest 4x3 and then |
| 148 | // crop; Otherwise, we find what the application asks for. Note that we assume |
| 149 | // that for HD, the desired format is always 16x9. The subclasses can override |
| 150 | // the default implementation. |
| 151 | // Parameters |
| 152 | // desired: the input desired format. If desired.fourcc is not kAnyFourcc, |
| 153 | // the best capture format has the exactly same fourcc. Otherwise, |
| 154 | // the best capture format uses a fourcc in GetPreferredFourccs(). |
| 155 | // best_format: the output of the best capture format. |
| 156 | // Return false if there is no such a best format, that is, the desired format |
| 157 | // is not supported. |
| 158 | virtual bool GetBestCaptureFormat(const VideoFormat& desired, |
| 159 | VideoFormat* best_format); |
| 160 | |
| 161 | // TODO(hellner): deprecate (make private) the Start API in favor of this one. |
| 162 | // Also remove CS_STARTING as it is implied by the return |
| 163 | // value of StartCapturing(). |
| 164 | bool StartCapturing(const VideoFormat& capture_format); |
| 165 | // Start the video capturer with the specified capture format. |
| 166 | // Parameter |
| 167 | // capture_format: The caller got this parameter by either calling |
| 168 | // GetSupportedFormats() and selecting one of the supported |
| 169 | // or calling GetBestCaptureFormat(). |
| 170 | // Return |
| 171 | // CS_STARTING: The capturer is trying to start. Success or failure will |
| 172 | // be notified via the |SignalStateChange| callback. |
| 173 | // CS_RUNNING: if the capturer is started and capturing. |
| 174 | // CS_PAUSED: Will never be returned. |
| 175 | // CS_FAILED: if the capturer failes to start.. |
| 176 | // CS_NO_DEVICE: if the capturer has no device and fails to start. |
| 177 | virtual CaptureState Start(const VideoFormat& capture_format) = 0; |
| 178 | // Sets the desired aspect ratio. If the capturer is capturing at another |
| 179 | // aspect ratio it will crop the width or the height so that asked for |
| 180 | // aspect ratio is acheived. Note that ratio_w and ratio_h do not need to be |
| 181 | // relatively prime. |
| 182 | void UpdateAspectRatio(int ratio_w, int ratio_h); |
| 183 | void ClearAspectRatio(); |
| 184 | |
| 185 | // Get the current capture format, which is set by the Start() call. |
| 186 | // Note that the width and height of the captured frames may differ from the |
| 187 | // capture format. For example, the capture format is HD but the captured |
| 188 | // frames may be smaller than HD. |
| 189 | const VideoFormat* GetCaptureFormat() const { |
| 190 | return capture_format_.get(); |
| 191 | } |
| 192 | |
| 193 | // Pause the video capturer. |
| 194 | virtual bool Pause(bool paused); |
| 195 | // Stop the video capturer. |
| 196 | virtual void Stop() = 0; |
| 197 | // Check if the video capturer is running. |
| 198 | virtual bool IsRunning() = 0; |
| 199 | // Restart the video capturer with the new |capture_format|. |
| 200 | // Default implementation stops and starts the capturer. |
| 201 | virtual bool Restart(const VideoFormat& capture_format); |
| 202 | // TODO(thorcarpenter): This behavior of keeping the camera open just to emit |
| 203 | // black frames is a total hack and should be fixed. |
| 204 | // When muting, produce black frames then pause the camera. |
| 205 | // When unmuting, start the camera. Camera starts unmuted. |
| 206 | virtual bool MuteToBlackThenPause(bool muted); |
| 207 | virtual bool IsMuted() const { |
| 208 | return muted_; |
| 209 | } |
| 210 | CaptureState capture_state() const { |
| 211 | return capture_state_; |
| 212 | } |
| 213 | |
| 214 | // Adds a video processor that will be applied on VideoFrames returned by |
| 215 | // |SignalVideoFrame|. Multiple video processors can be added. The video |
| 216 | // processors will be applied in the order they were added. |
| 217 | void AddVideoProcessor(VideoProcessor* video_processor); |
| 218 | // Removes the |video_processor| from the list of video processors or |
| 219 | // returns false. |
| 220 | bool RemoveVideoProcessor(VideoProcessor* video_processor); |
| 221 | |
| 222 | // Returns true if the capturer is screencasting. This can be used to |
| 223 | // implement screencast specific behavior. |
| 224 | virtual bool IsScreencast() const = 0; |
| 225 | |
| 226 | // Caps the VideoCapturer's format according to max_format. It can e.g. be |
| 227 | // used to prevent cameras from capturing at a resolution or framerate that |
| 228 | // the capturer is capable of but not performing satisfactorily at. |
| 229 | // The capping is an upper bound for each component of the capturing format. |
| 230 | // The fourcc component is ignored. |
| 231 | void ConstrainSupportedFormats(const VideoFormat& max_format); |
| 232 | |
| 233 | void set_enable_camera_list(bool enable_camera_list) { |
| 234 | enable_camera_list_ = enable_camera_list; |
| 235 | } |
| 236 | bool enable_camera_list() { |
| 237 | return enable_camera_list_; |
| 238 | } |
| 239 | // Signal all capture state changes that are not a direct result of calling |
| 240 | // Start(). |
| 241 | sigslot::signal2<VideoCapturer*, CaptureState> SignalStateChange; |
| 242 | // TODO(hellner): rename |SignalFrameCaptured| to something like |
| 243 | // |SignalRawFrame| or |SignalNativeFrame|. |
| 244 | // Frame callbacks are multithreaded to allow disconnect and connect to be |
| 245 | // called concurrently. It also ensures that it is safe to call disconnect |
| 246 | // at any time which is needed since the signal may be called from an |
| 247 | // unmarshalled thread owned by the VideoCapturer. |
| 248 | // Signal the captured frame to downstream. |
| 249 | sigslot::signal2<VideoCapturer*, const CapturedFrame*, |
| 250 | sigslot::multi_threaded_local> SignalFrameCaptured; |
| 251 | // Signal the captured frame converted to I420 to downstream. |
| 252 | sigslot::signal2<VideoCapturer*, const VideoFrame*, |
| 253 | sigslot::multi_threaded_local> SignalVideoFrame; |
| 254 | |
| 255 | const VideoProcessors& video_processors() const { return video_processors_; } |
| 256 | |
| 257 | protected: |
| 258 | // Callback attached to SignalFrameCaptured where SignalVideoFrames is called. |
| 259 | void OnFrameCaptured(VideoCapturer* video_capturer, |
| 260 | const CapturedFrame* captured_frame); |
| 261 | void SetCaptureState(CaptureState state); |
| 262 | |
| 263 | // Marshals SignalStateChange onto thread_. |
| 264 | void OnMessage(talk_base::Message* message); |
| 265 | |
| 266 | // subclasses override this virtual method to provide a vector of fourccs, in |
| 267 | // order of preference, that are expected by the media engine. |
| 268 | virtual bool GetPreferredFourccs(std::vector<uint32>* fourccs) = 0; |
| 269 | |
| 270 | // mutators to set private attributes |
| 271 | void SetId(const std::string& id) { |
| 272 | id_ = id; |
| 273 | } |
| 274 | |
| 275 | void SetCaptureFormat(const VideoFormat* format) { |
| 276 | capture_format_.reset(format ? new VideoFormat(*format) : NULL); |
| 277 | } |
| 278 | |
| 279 | void SetSupportedFormats(const std::vector<VideoFormat>& formats); |
| 280 | |
| 281 | private: |
| 282 | void Construct(); |
| 283 | // Get the distance between the desired format and the supported format. |
| 284 | // Return the max distance if they mismatch. See the implementation for |
| 285 | // details. |
| 286 | int64 GetFormatDistance(const VideoFormat& desired, |
| 287 | const VideoFormat& supported); |
| 288 | |
| 289 | // Convert captured frame to readable string for LOG messages. |
| 290 | std::string ToString(const CapturedFrame* frame) const; |
| 291 | |
| 292 | // Applies all registered processors. If any of the processors signal that |
| 293 | // the frame should be dropped the return value will be false. Note that |
| 294 | // this frame should be dropped as it has not applied all processors. |
| 295 | bool ApplyProcessors(VideoFrame* video_frame); |
| 296 | |
| 297 | // Updates filtered_supported_formats_ so that it contains the formats in |
| 298 | // supported_formats_ that fulfill all applied restrictions. |
| 299 | void UpdateFilteredSupportedFormats(); |
| 300 | // Returns true if format doesn't fulfill all applied restrictions. |
| 301 | bool ShouldFilterFormat(const VideoFormat& format) const; |
| 302 | |
| 303 | talk_base::Thread* thread_; |
| 304 | std::string id_; |
| 305 | CaptureState capture_state_; |
| 306 | talk_base::scoped_ptr<VideoFormat> capture_format_; |
| 307 | std::vector<VideoFormat> supported_formats_; |
| 308 | talk_base::scoped_ptr<VideoFormat> max_format_; |
| 309 | std::vector<VideoFormat> filtered_supported_formats_; |
| 310 | |
| 311 | int ratio_w_; // View resolution. e.g. 1280 x 720. |
| 312 | int ratio_h_; |
| 313 | bool enable_camera_list_; |
| 314 | int scaled_width_; // Current output size from ComputeScale. |
| 315 | int scaled_height_; |
| 316 | bool muted_; |
| 317 | int black_frame_count_down_; |
| 318 | |
| 319 | talk_base::CriticalSection crit_; |
| 320 | VideoProcessors video_processors_; |
| 321 | |
| 322 | DISALLOW_COPY_AND_ASSIGN(VideoCapturer); |
| 323 | }; |
| 324 | |
| 325 | } // namespace cricket |
| 326 | |
| 327 | #endif // TALK_MEDIA_BASE_VIDEOCAPTURER_H_ |