MTK|MTK Camera学习第四篇（拍照流程） MTKCamera学习第四篇（拍照流程

本篇仅学习从应用层到framework的过程，jni以下部分暂不讨论。因为一个优秀的相机应用，核心永远是它的图像处理部分（即Hal层中的3A算法部分），而MTK相关内容未开源。我们从按下拍照键开始做时序图如下：

文章图片
camera-takephoto.png
上图可以分成三个部分，第一部分就是接口的回调与实现，第二部分是不同模块execute执行指令部分，第三个部分是数据保存生成jpeg文件。我们这里说一下时序图上未记录的一些东西：
PhotoActor.java

@Override public void onShutterButtonClick(ShutterButton button) { int cameraState = mCameraActivity.getCameraState(); ViewState currentViewState = mICameraAppUi.getViewState(); Log.i(TAG, "[onShutterButtonClick] cameraState = " + cameraState + ", currentViewState = " + currentViewState); if (ViewState.VIEW_STATE_LOMOEFFECT_SETTING == currentViewState || ViewState.VIEW_STATE_CONTINUOUS_CAPTURE == currentViewState || ViewState.VIEW_STATE_CAMERA_CLOSED == currentViewState) { return; }if (mICameraAppUi.updateRemainStorage() > 0) { if (!(CameraActivity.STATE_SWITCHING_CAMERA == cameraState || CameraActivity.STATE_PREVIEW_STOPPED == cameraState)) { if (mSelfTimerManager.startSelfTimer()) { Log.i(TAG, "[onShutterButtonClick] start self timer"); mModuleManager.onSelfTimerState(true); mICameraAppUi.setSwipeEnabled(false); mICameraAppUi.setViewState(ViewState.VIEW_STATE_CAPTURE); mIsSelftimerCounting = true; return; } else { mIsSelftimerCounting = false; }mModuleManager.onPhotoShutterButtonClick(); } } else { Log.i(TAG, "remain storage is less than 0"); mICameraAppUi.showRemaining(); } }

这里面有两个函数未记入时序图，一个updateRemainStorage()，其最终实现位于RemainingManager用来计算当前的剩余空间，空间不足自然无法拍照保存，另一个是mSelfTimerManager.startSelfTimer()用来做倒计时拍照的功能。
ModuleManager.java

public boolean onPhotoShutterButtonClick() { mAdditionManager.execute(ActionType.ACTION_PHOTO_SHUTTER_BUTTON_CLICK, false); return mICameraMode.execute(ActionType.ACTION_PHOTO_SHUTTER_BUTTON_CLICK); }

在这个地方开始执行execute，mAdditionManager是什么呢？我们进入这个类，看一下它的execute方法：

public boolean execute(AdditionActionType type, Object... arg) { Log.i(TAG, "[execute],addition action type = " + type); boolean result = false; for (ICameraAddition addition : mModeAddition) { result = addition.execute(type, arg) || result; } return result; }

mModeAddition表示当前的拍摄模式，其值定义如下：

public void setCurrentMode(CameraModeType type) { Log.i(TAG, "[setCurrentMode]type = " + type); switch (type) { case EXT_MODE_PHOTO: mModeAddition = mPhotoAddtion; break; case EXT_MODE_VIDEO: mModeAddition = mVideoAddtion; break; case EXT_MODE_PHOTO_PIP: mModeAddition = mPipPhotoAddition; break; case EXT_MODE_VIDEO_PIP: mModeAddition = mPipVideoAddition; break; case EXT_MODE_FACE_BEAUTY: mModeAddition = mFaceBeautyAddition; break; case EXT_MODE_STEREO_CAMERA: mModeAddition = mRefocusAddition; break; default: mModeAddition = mDummyAddtion; break; } }

我们从相关各个类的关系来看下：

文章图片
camera-addition.png
CameraAddition的子类除了图中的还有其它，这里只画出了在AdditionManager初始化时用到的子类。整个addition包是由mtk新加的code，从ICameraAddition该接口的定义来看，大概是将不同的ACTION_TYPE传递到不同的MODE下进行Camera的操作。
回到前面mICameraMode.execute()，根据不同的mode进入相关类，我们这里肯定是执行PhotoMode的execute方法。顺序执行到如下方法：

private void onShutterButtonClick() { boolean isEnoughSpace = mIFileSaver.isEnoughSpace(); Log.i(TAG, "[onShutterButtonClick]isEnoughSpace = " + isEnoughSpace + ",mCameraClosed = " + mCameraClosed + ",mCurrentState = " + getModeState()); // Do not take the picture if there is not enough storage or camera is not available. if (!isEnoughSpace || isCameraNotAvailable()) { Log.w(TAG, "[onShutterButtonClick]return."); return; } Log.i(TAG, "[CMCC Performance test][Camera][Camera] camera capture start [" + System.currentTimeMillis() + "]"); if (mIFocusManager != null) { mIFocusManager.focusAndCapture(); } }

这里FileSaverImpl会判断一次剩余空间的大小，在前面第一次判断剩余空间时，将avaliableSpace保存到Storage，而在这里会将该值取出再次判断。而当相机处于不可用状态时也将放弃拍照而直接返回。

private boolean isCameraNotAvailable() { ModeState modeState = getModeState(); Log.d(TAG, "isCameraNotAvailable modeState " + modeState); return (ModeState.STATE_CAPTURING == modeState || ModeState.STATE_SAVING == modeState || ModeState.STATE_CLOSED == modeState) ? true : false; }

准确的来说，应该不是相机不可用，而是当相机处于CAPTURING、SAVING、CLOSED这三种状态时暂时看作相机不可用。如果不是这三种状态，则开始聚焦拍照。然后，接口回调到PhotoMode中的capture()方法：

@Override public boolean capture() { Log.i(TAG, "[capture]..."); long start = System.currentTimeMillis(); mCaptureStartTime = System.currentTimeMillis(); mPostViewPictureCallbackTime = 0; mJpegImageData = https://www.it610.com/article/null; mIFileSaver.init(FILE_TYPE.JPEG, 0, null, -1); mICameraAppUi.setSwipeEnabled(false); mICameraAppUi.showRemaining(); mCameraCategory.takePicture(); setModeState(ModeState.STATE_CAPTURING); Log.d(TAG,"[capture] Capture time = " + (System.currentTimeMillis() - start)); return true; }

这里各种工作准备就绪后初始化FileSaver相关准备进行文件保存，然后更新UI显示当前的剩余空间，然后调用了内部类的方法再通过代理调用到了Camera。

protected class CameraCategory {public void takePicture() { if (!mAdditionManager.execute(AdditionActionType.ACTION_TAKEN_PICTURE)) { mICameraDevice.takePicture(mShutterCallback, mRawPictureCallback, null, mJpegPictureCallback); mICameraAppUi.setViewState(ViewState.VIEW_STATE_CAPTURE); } } }

我们的文件保存在哪进行的呢，看下这个mJpegPictureCallback：

private final PictureCallback mJpegPictureCallback = new PictureCallback() { @Override public void onPictureTaken(byte[] jpegData, Camera camera) { Log.i(TAG, "[mJpegPictureCallback]onPictureTaken"); ... // Calculate the width and the height of the jpeg. if (!mIModuleCtrl.isImageCaptureIntent()) { mIFileSaver.savePhotoFile(jpegData, null, mCaptureStartTime, mIModuleCtrl.getLocation(), 0, null); } else { mJpegImageData = https://www.it610.com/article/jpegData; if (!mIModuleCtrl.isQuickCapture()) { mICameraAppUi.showReview(null, null); mICameraAppUi.switchShutterType(ShutterButtonType.SHUTTER_TYPE_OK_CANCEL); } else { doAttach(); } } ... } };

SaveRequest将保存照片的各种信息：

@Override public boolean savePhotoFile(byte[] photoData, String fileName, long date, Location location, int tag, OnFileSavedListener listener) { Log.i(TAG, "[savePhotoFile]title =" + fileName); if (null == mSaveRequest || null == photoData) { Log.w(TAG, "[savePhotoFile]fail,mSaveRequest = " + mSaveRequest); return false; }mListener = listener; if (mSaveRequest.getDataSize() > 0) { Log.i(TAG, "[savePhotoFile]Current SaveRequest is used, copy new one!"); mSaveRequest = mFileSaver.copyPhotoRequest(mSaveRequest); } mSaveRequest.setData(photoData); mSaveRequest.setFileName(fileName); mSaveRequest.setTag(tag); mSaveRequest.updateDataTaken(date); mSaveRequest.setLocation(location); mSaveRequest.setListener(mFileSaverListener); mSaveRequest.addRequest(); return true; }

SaveRequest是一个接口，其实现在FileSaver的内部类RequestOperator，而该内部类同时又是一个抽象类，其子类同样是位于FileSaver的内部类VideoOperator、PanoOperator、PhotoOperator，因此这里的addRequet方法最终由PhotoOperator实现，它又将调用外部类FileSaver的addSaveRequest()方法，而这个请求在这里将交给FileSaverService来处理：

// run in main thread public void addSaveRequest(SaveRequest request) { Log.i(TAG, "[addSaveRequest]...begin,the queue number is = " + mQueue.size() + "mContinuousSaveTask:" + mContinuousSaveTask); synchronized (mQueue) { mQueue.add(request); } if (mContinuousSaveTask == null) { mContinuousSaveTask = new SaveTask(); mTaskNumber++; mContinuousSaveTask.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR); Log.i(TAG, "[addSaveRequest]execute continuous AsyncTask."); } Log.i(TAG, "[addSaveRequest]...end,the queue number is = " + mQueue.size()); }

因为需要写文件，这里开启了一个AsyncTask来处理耗时操作：

@Override protected Void doInBackground(Void... v) { Log.i(TAG, "[SaveTask]doInBackground...,queue is empty = " + mQueue.isEmpty()); FileSaverListener lastFileSaverListener = null; while (!mQueue.isEmpty()) { r = mQueue.get(0); //different cameraActivity use different listener // notify old listener save done info if (lastFileSaverListener != null && r.getFileSaverListener() != lastFileSaverListener) { r.getFileSaverListener().onSaveDone(); }if (Storage.isStorageReady()) { r.saveRequest(); } r.notifyListener(); // LinkedList is not saved list, so mQueue should be sync in // multi thread synchronized (mQueue) { mQueue.remove(0); } synchronized (mListnerObject) { r.getFileSaverListener().onFileSaved(r); } lastFileSaverListener = r.getFileSaverListener(); } mContinuousSaveTask = null; mTaskNumber--; synchronized (mListnerObject) { r.getFileSaverListener().onSaveDone(); } Log.i(TAG, "[SaveTask]doInBackground...,end "); return null; }

通知相关Listener保存成功。SD卡准备好的话，具体的写文件操作重新交给SaveRequest（由java特性确定最后由PhotoOperator实现）

@Override public synchronized void saveRequest() { if (mData =https://www.it610.com/article/= null) { Log.w(TAG,"[saveRequest]mData is null,return!"); return; } int orientation = Exif.getOrientation(mData); // M: ConShots mGroupId = Exif.getGroupId(mData); mGroupIndex = Exif.getGroupIndex(mData); mFocusValueHigh = Exif.getFocusValueHigh(mData); mFocusValueLow = Exif.getFocusValueLow(mData); mOrientation = orientation; mDataSize = mData.length; if (null != mFileName) { mTitle = mFileName.substring(0, mFileName.indexOf('.')); } else { mTitle = createName(mFileType, mDateTaken, mGroupIndex); mFileName = Storage.generateFileName(mTitle, mTempPictureType); Log.i(TAG, "[saveRequest]PhotoOperator,mFileName = " + mFileName); } mFilePath = Storage.generateFilepath(mFileName); mTempFilePath = mFilePath + TEMP_SUFFIX; saveImageToSDCard(mTempFilePath, mFilePath, mData); // camera decouple mMimeType = Storage.generateMimetype(mTitle, mTempPictureType); checkDataProperty(); saveImageToDatabase(this); } private void saveImageToSDCard(String tempFilePath, String filePath, byte[] data) { FileOutputStream out = null; try { // Write to a temporary file and rename it to the final name. // This // avoids other apps reading incomplete data. Log.d(TAG, "[saveImageToSDCard]begin add the data to SD Card"); out = new FileOutputStream(tempFilePath); out.write(data); out.close(); new File(tempFilePath).renameTo(new File(filePath)); } catch (IOException e) { Log.e(TAG, "[saveImageToSDCard]Failed to write image,ex:", e); } finally { if (out != null) { try { out.close(); } catch (IOException e) { Log.e(TAG, "[saveImageToSDCard]IOException:", e); } } } Log.i(TAG, "[saveImageToSDCard]end of add the data to SD Card"); }

我们看到最后的文件生成就是一个new File的过程，这说明我们在PictureCallback里面拿到的数据一定是已经过3A算法处理后的数据，而不是Camera设备得到的原始数据。最后我们看下FileSaver相关类的关系图：

文章图片
camera-filesaver.png
【MTK|MTK Camera学习第四篇（拍照流程）】为方便看图，这里没有加入VideoOperator和PanoOperator这两个内部类。我们在PictureCallback中将数据保存为文件的时候，直接操作的是IFileSaver这个接口，其实现是FileSaverImpl，然后该类中有两个成员变量mSaveRequest和mVideoSaveRequest（接口SaveRequest的对象），执行它们的addRequest()方法也就是执行FileSaver的内部类RequestOperator的想着方法，而由于其抽象类型的原因最终又将来到其子类PhotoOperator的相关方法，与我们上面的时序图一致。