OLD | NEW |
| (Empty) |
1 /* | |
2 * libjingle | |
3 * Copyright 2010 Google Inc. | |
4 * | |
5 * Redistribution and use in source and binary forms, with or without | |
6 * modification, are permitted provided that the following conditions are met: | |
7 * | |
8 * 1. Redistributions of source code must retain the above copyright notice, | |
9 * this list of conditions and the following disclaimer. | |
10 * 2. Redistributions in binary form must reproduce the above copyright notice, | |
11 * this list of conditions and the following disclaimer in the documentation | |
12 * and/or other materials provided with the distribution. | |
13 * 3. The name of the author may not be used to endorse or promote products | |
14 * derived from this software without specific prior written permission. | |
15 * | |
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED | |
17 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF | |
18 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO | |
19 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
20 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, | |
21 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; | |
22 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, | |
23 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR | |
24 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF | |
25 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
26 */ | |
27 | |
28 // Declaration of abstract class VideoCapturer | |
29 | |
30 #ifndef TALK_MEDIA_BASE_VIDEOCAPTURER_H_ | |
31 #define TALK_MEDIA_BASE_VIDEOCAPTURER_H_ | |
32 | |
33 #include <algorithm> | |
34 #include <string> | |
35 #include <vector> | |
36 | |
37 #include "talk/media/base/mediachannel.h" | |
38 #include "talk/media/base/videoadapter.h" | |
39 #include "talk/media/base/videocommon.h" | |
40 #include "talk/media/base/videoframefactory.h" | |
41 #include "talk/media/devices/devicemanager.h" | |
42 #include "webrtc/base/basictypes.h" | |
43 #include "webrtc/base/criticalsection.h" | |
44 #include "webrtc/base/messagehandler.h" | |
45 #include "webrtc/base/rollingaccumulator.h" | |
46 #include "webrtc/base/scoped_ptr.h" | |
47 #include "webrtc/base/sigslot.h" | |
48 #include "webrtc/base/thread.h" | |
49 #include "webrtc/base/timing.h" | |
50 | |
51 | |
52 namespace cricket { | |
53 | |
54 // Current state of the capturer. | |
55 // TODO(hellner): CS_NO_DEVICE is an error code not a capture state. Separate | |
56 // error codes and states. | |
57 enum CaptureState { | |
58 CS_STOPPED, // The capturer has been stopped or hasn't started yet. | |
59 CS_STARTING, // The capturer is in the process of starting. Note, it may | |
60 // still fail to start. | |
61 CS_RUNNING, // The capturer has been started successfully and is now | |
62 // capturing. | |
63 CS_PAUSED, // The capturer has been paused. | |
64 CS_FAILED, // The capturer failed to start. | |
65 CS_NO_DEVICE, // The capturer has no device and consequently failed to start. | |
66 }; | |
67 | |
68 class VideoFrame; | |
69 | |
70 struct CapturedFrame { | |
71 static const uint32_t kFrameHeaderSize = 40; // Size from width to data_size. | |
72 static const uint32_t kUnknownDataSize = 0xFFFFFFFF; | |
73 | |
74 CapturedFrame(); | |
75 | |
76 // Get the number of bytes of the frame data. If data_size is known, return | |
77 // it directly. Otherwise, calculate the size based on width, height, and | |
78 // fourcc. Return true if succeeded. | |
79 bool GetDataSize(uint32_t* size) const; | |
80 | |
81 // The width and height of the captured frame could be different from those | |
82 // of VideoFormat. Once the first frame is captured, the width, height, | |
83 // fourcc, pixel_width, and pixel_height should keep the same over frames. | |
84 int width; // in number of pixels | |
85 int height; // in number of pixels | |
86 uint32_t fourcc; // compression | |
87 uint32_t pixel_width; // width of a pixel, default is 1 | |
88 uint32_t pixel_height; // height of a pixel, default is 1 | |
89 int64_t time_stamp; // timestamp of when the frame was captured, in unix | |
90 // time with nanosecond units. | |
91 uint32_t data_size; // number of bytes of the frame data | |
92 | |
93 webrtc::VideoRotation rotation; // rotation in degrees of the frame. | |
94 | |
95 void* data; // pointer to the frame data. This object allocates the | |
96 // memory or points to an existing memory. | |
97 | |
98 private: | |
99 RTC_DISALLOW_COPY_AND_ASSIGN(CapturedFrame); | |
100 }; | |
101 | |
102 // VideoCapturer is an abstract class that defines the interfaces for video | |
103 // capturing. The subclasses implement the video capturer for various types of | |
104 // capturers and various platforms. | |
105 // | |
106 // The captured frames may need to be adapted (for example, cropping). | |
107 // Video adaptation is built into and enabled by default. After a frame has | |
108 // been captured from the device, it is sent to the video adapter, then out to | |
109 // the encoder. | |
110 // | |
111 // Programming model: | |
112 // Create an object of a subclass of VideoCapturer | |
113 // Initialize | |
114 // SignalStateChange.connect() | |
115 // SignalFrameCaptured.connect() | |
116 // Find the capture format for Start() by either calling GetSupportedFormats() | |
117 // and selecting one of the supported or calling GetBestCaptureFormat(). | |
118 // video_adapter()->OnOutputFormatRequest(desired_encoding_format) | |
119 // Start() | |
120 // GetCaptureFormat() optionally | |
121 // Stop() | |
122 // | |
123 // Assumption: | |
124 // The Start() and Stop() methods are called by a single thread (E.g., the | |
125 // media engine thread). Hence, the VideoCapture subclasses dont need to be | |
126 // thread safe. | |
127 // | |
128 class VideoCapturer | |
129 : public sigslot::has_slots<>, | |
130 public rtc::MessageHandler { | |
131 public: | |
132 // All signals are marshalled to |thread| or the creating thread if | |
133 // none is provided. | |
134 VideoCapturer(); | |
135 explicit VideoCapturer(rtc::Thread* thread); | |
136 virtual ~VideoCapturer() {} | |
137 | |
138 // Gets the id of the underlying device, which is available after the capturer | |
139 // is initialized. Can be used to determine if two capturers reference the | |
140 // same device. | |
141 const std::string& GetId() const { return id_; } | |
142 | |
143 // Get the capture formats supported by the video capturer. The supported | |
144 // formats are non empty after the device has been opened successfully. | |
145 const std::vector<VideoFormat>* GetSupportedFormats() const; | |
146 | |
147 // Get the best capture format for the desired format. The best format is the | |
148 // same as one of the supported formats except that the frame interval may be | |
149 // different. If the application asks for 16x9 and the camera does not support | |
150 // 16x9 HD or the application asks for 16x10, we find the closest 4x3 and then | |
151 // crop; Otherwise, we find what the application asks for. Note that we assume | |
152 // that for HD, the desired format is always 16x9. The subclasses can override | |
153 // the default implementation. | |
154 // Parameters | |
155 // desired: the input desired format. If desired.fourcc is not kAnyFourcc, | |
156 // the best capture format has the exactly same fourcc. Otherwise, | |
157 // the best capture format uses a fourcc in GetPreferredFourccs(). | |
158 // best_format: the output of the best capture format. | |
159 // Return false if there is no such a best format, that is, the desired format | |
160 // is not supported. | |
161 virtual bool GetBestCaptureFormat(const VideoFormat& desired, | |
162 VideoFormat* best_format); | |
163 | |
164 // TODO(hellner): deprecate (make private) the Start API in favor of this one. | |
165 // Also remove CS_STARTING as it is implied by the return | |
166 // value of StartCapturing(). | |
167 bool StartCapturing(const VideoFormat& capture_format); | |
168 // Start the video capturer with the specified capture format. | |
169 // Parameter | |
170 // capture_format: The caller got this parameter by either calling | |
171 // GetSupportedFormats() and selecting one of the supported | |
172 // or calling GetBestCaptureFormat(). | |
173 // Return | |
174 // CS_STARTING: The capturer is trying to start. Success or failure will | |
175 // be notified via the |SignalStateChange| callback. | |
176 // CS_RUNNING: if the capturer is started and capturing. | |
177 // CS_PAUSED: Will never be returned. | |
178 // CS_FAILED: if the capturer failes to start.. | |
179 // CS_NO_DEVICE: if the capturer has no device and fails to start. | |
180 virtual CaptureState Start(const VideoFormat& capture_format) = 0; | |
181 // Sets the desired aspect ratio. If the capturer is capturing at another | |
182 // aspect ratio it will crop the width or the height so that asked for | |
183 // aspect ratio is acheived. Note that ratio_w and ratio_h do not need to be | |
184 // relatively prime. | |
185 void UpdateAspectRatio(int ratio_w, int ratio_h); | |
186 void ClearAspectRatio(); | |
187 | |
188 // Get the current capture format, which is set by the Start() call. | |
189 // Note that the width and height of the captured frames may differ from the | |
190 // capture format. For example, the capture format is HD but the captured | |
191 // frames may be smaller than HD. | |
192 const VideoFormat* GetCaptureFormat() const { | |
193 return capture_format_.get(); | |
194 } | |
195 | |
196 // Pause the video capturer. | |
197 virtual bool Pause(bool paused); | |
198 // Stop the video capturer. | |
199 virtual void Stop() = 0; | |
200 // Check if the video capturer is running. | |
201 virtual bool IsRunning() = 0; | |
202 // Restart the video capturer with the new |capture_format|. | |
203 // Default implementation stops and starts the capturer. | |
204 virtual bool Restart(const VideoFormat& capture_format); | |
205 // TODO(thorcarpenter): This behavior of keeping the camera open just to emit | |
206 // black frames is a total hack and should be fixed. | |
207 // When muting, produce black frames then pause the camera. | |
208 // When unmuting, start the camera. Camera starts unmuted. | |
209 virtual bool MuteToBlackThenPause(bool muted); | |
210 virtual bool IsMuted() const { | |
211 return muted_; | |
212 } | |
213 CaptureState capture_state() const { | |
214 return capture_state_; | |
215 } | |
216 | |
217 // Tells videocapturer whether to apply the pending rotation. By default, the | |
218 // rotation is applied and the generated frame is up right. When set to false, | |
219 // generated frames will carry the rotation information from | |
220 // SetCaptureRotation. Return value indicates whether this operation succeeds. | |
221 virtual bool SetApplyRotation(bool enable); | |
222 virtual bool GetApplyRotation() { return apply_rotation_; } | |
223 | |
224 // Returns true if the capturer is screencasting. This can be used to | |
225 // implement screencast specific behavior. | |
226 virtual bool IsScreencast() const = 0; | |
227 | |
228 // Caps the VideoCapturer's format according to max_format. It can e.g. be | |
229 // used to prevent cameras from capturing at a resolution or framerate that | |
230 // the capturer is capable of but not performing satisfactorily at. | |
231 // The capping is an upper bound for each component of the capturing format. | |
232 // The fourcc component is ignored. | |
233 void ConstrainSupportedFormats(const VideoFormat& max_format); | |
234 | |
235 void set_enable_camera_list(bool enable_camera_list) { | |
236 enable_camera_list_ = enable_camera_list; | |
237 } | |
238 bool enable_camera_list() { | |
239 return enable_camera_list_; | |
240 } | |
241 | |
242 // Enable scaling to ensure square pixels. | |
243 void set_square_pixel_aspect_ratio(bool square_pixel_aspect_ratio) { | |
244 square_pixel_aspect_ratio_ = square_pixel_aspect_ratio; | |
245 } | |
246 bool square_pixel_aspect_ratio() { | |
247 return square_pixel_aspect_ratio_; | |
248 } | |
249 | |
250 // Signal all capture state changes that are not a direct result of calling | |
251 // Start(). | |
252 sigslot::signal2<VideoCapturer*, CaptureState> SignalStateChange; | |
253 // Frame callbacks are multithreaded to allow disconnect and connect to be | |
254 // called concurrently. It also ensures that it is safe to call disconnect | |
255 // at any time which is needed since the signal may be called from an | |
256 // unmarshalled thread owned by the VideoCapturer. | |
257 // Signal the captured frame to downstream. | |
258 sigslot::signal2<VideoCapturer*, const CapturedFrame*, | |
259 sigslot::multi_threaded_local> SignalFrameCaptured; | |
260 // Signal the captured and possibly adapted frame to downstream consumers | |
261 // such as the encoder. | |
262 sigslot::signal2<VideoCapturer*, const VideoFrame*, | |
263 sigslot::multi_threaded_local> SignalVideoFrame; | |
264 | |
265 // If true, run video adaptation. By default, video adaptation is enabled | |
266 // and users must call video_adapter()->OnOutputFormatRequest() | |
267 // to receive frames. | |
268 bool enable_video_adapter() const { return enable_video_adapter_; } | |
269 void set_enable_video_adapter(bool enable_video_adapter) { | |
270 enable_video_adapter_ = enable_video_adapter; | |
271 } | |
272 | |
273 CoordinatedVideoAdapter* video_adapter() { return &video_adapter_; } | |
274 const CoordinatedVideoAdapter* video_adapter() const { | |
275 return &video_adapter_; | |
276 } | |
277 | |
278 // Takes ownership. | |
279 void set_frame_factory(VideoFrameFactory* frame_factory); | |
280 | |
281 // Gets statistics for tracked variables recorded since the last call to | |
282 // GetStats. Note that calling GetStats resets any gathered data so it | |
283 // should be called only periodically to log statistics. | |
284 void GetStats(VariableInfo<int>* adapt_drop_stats, | |
285 VariableInfo<int>* effect_drop_stats, | |
286 VariableInfo<double>* frame_time_stats, | |
287 VideoFormat* last_captured_frame_format); | |
288 | |
289 protected: | |
290 // Callback attached to SignalFrameCaptured where SignalVideoFrames is called. | |
291 void OnFrameCaptured(VideoCapturer* video_capturer, | |
292 const CapturedFrame* captured_frame); | |
293 void SetCaptureState(CaptureState state); | |
294 | |
295 // Marshals SignalStateChange onto thread_. | |
296 void OnMessage(rtc::Message* message); | |
297 | |
298 // subclasses override this virtual method to provide a vector of fourccs, in | |
299 // order of preference, that are expected by the media engine. | |
300 virtual bool GetPreferredFourccs(std::vector<uint32_t>* fourccs) = 0; | |
301 | |
302 // mutators to set private attributes | |
303 void SetId(const std::string& id) { | |
304 id_ = id; | |
305 } | |
306 | |
307 void SetCaptureFormat(const VideoFormat* format) { | |
308 capture_format_.reset(format ? new VideoFormat(*format) : NULL); | |
309 if (capture_format_) { | |
310 ASSERT(capture_format_->interval > 0 && | |
311 "Capture format expected to have positive interval."); | |
312 // Video adapter really only cares about capture format interval. | |
313 video_adapter_.SetInputFormat(*capture_format_); | |
314 } | |
315 } | |
316 | |
317 void SetSupportedFormats(const std::vector<VideoFormat>& formats); | |
318 VideoFrameFactory* frame_factory() { return frame_factory_.get(); } | |
319 | |
320 private: | |
321 void Construct(); | |
322 // Get the distance between the desired format and the supported format. | |
323 // Return the max distance if they mismatch. See the implementation for | |
324 // details. | |
325 int64_t GetFormatDistance(const VideoFormat& desired, | |
326 const VideoFormat& supported); | |
327 | |
328 // Convert captured frame to readable string for LOG messages. | |
329 std::string ToString(const CapturedFrame* frame) const; | |
330 | |
331 // Updates filtered_supported_formats_ so that it contains the formats in | |
332 // supported_formats_ that fulfill all applied restrictions. | |
333 void UpdateFilteredSupportedFormats(); | |
334 // Returns true if format doesn't fulfill all applied restrictions. | |
335 bool ShouldFilterFormat(const VideoFormat& format) const; | |
336 | |
337 void UpdateStats(const CapturedFrame* captured_frame); | |
338 | |
339 // Helper function to save statistics on the current data from a | |
340 // RollingAccumulator into stats. | |
341 template<class T> | |
342 static void GetVariableSnapshot( | |
343 const rtc::RollingAccumulator<T>& data, | |
344 VariableInfo<T>* stats); | |
345 | |
346 rtc::Thread* thread_; | |
347 std::string id_; | |
348 CaptureState capture_state_; | |
349 rtc::scoped_ptr<VideoFrameFactory> frame_factory_; | |
350 rtc::scoped_ptr<VideoFormat> capture_format_; | |
351 std::vector<VideoFormat> supported_formats_; | |
352 rtc::scoped_ptr<VideoFormat> max_format_; | |
353 std::vector<VideoFormat> filtered_supported_formats_; | |
354 | |
355 int ratio_w_; // View resolution. e.g. 1280 x 720. | |
356 int ratio_h_; | |
357 bool enable_camera_list_; | |
358 bool square_pixel_aspect_ratio_; // Enable scaling to square pixels. | |
359 int scaled_width_; // Current output size from ComputeScale. | |
360 int scaled_height_; | |
361 bool muted_; | |
362 int black_frame_count_down_; | |
363 | |
364 bool enable_video_adapter_; | |
365 CoordinatedVideoAdapter video_adapter_; | |
366 | |
367 rtc::Timing frame_length_time_reporter_; | |
368 rtc::CriticalSection frame_stats_crit_; | |
369 | |
370 int adapt_frame_drops_; | |
371 rtc::RollingAccumulator<int> adapt_frame_drops_data_; | |
372 double previous_frame_time_; | |
373 rtc::RollingAccumulator<double> frame_time_data_; | |
374 // The captured frame format before potential adapation. | |
375 VideoFormat last_captured_frame_format_; | |
376 | |
377 // Whether capturer should apply rotation to the frame before signaling it. | |
378 bool apply_rotation_; | |
379 | |
380 RTC_DISALLOW_COPY_AND_ASSIGN(VideoCapturer); | |
381 }; | |
382 | |
383 } // namespace cricket | |
384 | |
385 #endif // TALK_MEDIA_BASE_VIDEOCAPTURER_H_ | |
OLD | NEW |