我正在构建一个缓冲 N 相机帧的应用程序，当用户点击一个按钮时，它会使用应用效果的所有已保存帧来保存照片。
我在 AsyncTask 上保存照片并处理帧。当我执行它时，我从屏幕上删除了所有内容，只留下一个 TextView 来显示保存照片的进度。
目前 AsyncTask doInBackground 看起来像这样:
protected Void doInBackground(Integer... params)
{
int w = mBuffer.get(0).getWidth();
int h = mBuffer.get(0).getHeight();
int lineHeight = h / mBuffer.size();
int currentHeight = 0;
Log.d("output", "saving photo "+w+", "+h);
for (int i = 0; i < mBuffer.size(); i++)
{
YuvImage image = mBuffer.get(i);
ByteArrayOutputStream out = new ByteArrayOutputStream();
if (!image.compressToJpeg(new Rect(0, currentHeight, w, lineHeight), 50, out))
{
Log.d("output", "problem converting yuv to jpg");
break;
}
currentHeight += lineHeight;
Log.d("output", "currentHeight: "+currentHeight);
publishProgress((int)((i / (float)mBuffer.size()) * 100));
try
{
out.flush();
out.close();
out = null;
}
catch(Exception e)
{
}
System.gc();
try
{
Thread.sleep(200);
}
catch(Exception e)
{
}
if (isCancelled())
{
break;
}
}
return null;
}
我已经为这个问题删除了很多不太重要的代码部分，这只是引发 OutOfMemory 问题的部分。基本上我在那里做的是从图像中获取一行并将其压缩为 JPEG。但它只能压缩第一张图片。当它转到第二个时，它会引发一个 OutOfMemory 异常。
System.GC(), Thread.sleep(), out.flush(), out.close( )，都是解决问题的不成功尝试。
当前mBuffer的大小是5，最初是32。从DDMS Heap调试来看，它的Heap Size是30MB，正在分配9MB。显然还有很大的成长空间。如果我删除 compressToJpeg()，AsyncTask 就可以正常完成。
有人能解决这个问题吗？我应该尝试自己的 YUV -> JPEG 转换器吗？
使用最新代码进行编辑:
protected Void doInBackground(Integer... params)
{
int lineHeight = mHeight / mBufferSize;
int currentHeight = 0;
Log.d("output", "saving photo: "+mWidth+", "+mHeight);
File outputPath = new File(Environment.getExternalStorageDirectory().getPath() + "/camerafluid-output");
outputPath.mkdirs();
for (int i = 0; i < mBufferSize; i++)
{
File imageFile = new File(mCachePath, "image-"+0);
File outputFile = new File(outputPath, "image-"+0);
if (!imageFile.exists())
{
Log.d("output", "image "+i+" not found on cache directory");
continue;
}
try
{
int size = (int)imageFile.length();
byte[] imageBytes = new byte[size];
BufferedInputStream buf = new BufferedInputStream(new FileInputStream(imageFile));
buf.read(imageBytes, 0, size);
buf.close();
YuvImage image = new YuvImage(imageBytes, ImageFormat.NV21, mWidth, mHeight, null);
imageBytes = null;
ByteArrayOutputStream out = new ByteArrayOutputStream();
if (!image.compressToJpeg(new Rect(0, currentHeight, mWidth, lineHeight), 50, out))
{
Log.d("output", "problem converting yuv to jpg");
break;
}
FileOutputStream s = new FileOutputStream(outputFile);
s.write(out.toByteArray());
s.flush();
s.close();
s = null;
currentHeight += lineHeight;
Log.d("output", "currentHeight: "+currentHeight);
publishProgress((int)((i / (float)mBufferSize) * 100));
System.gc();
}
catch(Exception e)
{
}
if (isCancelled())
{
break;
}
}
return null;
}
我正确地保存了图像，因为我可以在缓存目录中看到保存的文件，这是保存缓冲区并执行 AsyncTask 的 Activity 部分:
public void takePhoto()
{
Log.d("output", "taking photo");
//vfranchi - save the buffer on the disk to free memory
File cachePath = new File(getExternalCacheDir().getPath() + "/buffer");
if (!cachePath.exists())
{
Log.d("output", "cache path doesnt exist");
cachePath.mkdirs();
}
int count = 0;
int width = mPhotoBuffer.get(0).getWidth();
int height = mPhotoBuffer.get(0).getHeight();
int bufferSize = mPhotoBuffer.size();
for(YuvImage i : mPhotoBuffer)
{
File f = new File(cachePath, "image-"+count);
try
{
FileOutputStream s = new FileOutputStream(f);
s.write(i.getYuvData());
s.flush();
s.close();
Log.d("output", "saved image "+count);
count++;
}
catch(Exception e)
{
Log.d("output", "unable to save image "+count+"\n"+e.getLocalizedMessage());
}
}
mPhotoBuffer = null;
System.gc();
SavePhotoTask task = new SavePhotoTask(this, cachePath, width, height, bufferSize);
task.execute(0);
}
Logcat 开始拍照方法:
03-30 12:48:08.534: D/output(15359): taking photo
03-30 12:48:09.011: V/Camera-JNI(15359): setHasPreviewCallback: installed:0, manualBuffer:0
03-30 12:48:09.011: V/Camera-JNI(15359): get_native_camera: context=0x138520, camera=0x1eea78
03-30 12:48:09.011: V/Camera-JNI(15359): Clearing callback buffers, 0 remained
03-30 12:48:09.011: V/Camera-JNI(15359): stopPreview
03-30 12:48:09.011: V/Camera-JNI(15359): get_native_camera: context=0x138520, camera=0x1eea78
03-30 12:48:15.323: D/output(15359): saved image 0
03-30 12:48:16.628: D/output(15359): saved image 1
03-30 12:48:17.880: D/output(15359): saved image 2
03-30 12:48:19.073: D/output(15359): saved image 3
03-30 12:48:20.545: D/output(15359): saved image 4
03-30 12:48:28.089: D/output(15359): saving photo: 848, 480
03-30 12:48:33.276: D/skia(15359): onFlyCompress
03-30 12:48:35.020: D/output(15359): currentHeight: 96
03-30 12:48:44.144: D/output(15359): currentHeight: 192
03-30 12:48:52.581: D/skia(15359): onFlyCompress
03-30 12:48:52.878: I/dalvikvm-heap(15359): Grow heap (frag case) to 14.975MB for 2095120-byte allocation
03-30 12:48:53.112: I/dalvikvm-heap(15359): Grow heap (frag case) to 17.975MB for 4192272-byte allocation
03-30 12:48:53.566: I/dalvikvm-heap(15359): Grow heap (frag case) to 23.975MB for 8386576-byte allocation
03-30 12:48:54.855: I/dalvikvm-heap(15359): Grow heap (frag case) to 35.975MB for 16775184-byte allocation
03-30 12:48:56.620: I/dalvikvm-heap(15359): Forcing collection of SoftReferences for 33552400-byte allocation
03-30 12:48:56.644: E/dalvikvm-heap(15359): Out of memory on a 33552400-byte allocation.
03-30 12:48:56.644: I/dalvikvm(15359): "AsyncTask #1" prio=5 tid=12 RUNNABLE
03-30 12:48:56.644: I/dalvikvm(15359): | group="main" sCount=0 dsCount=0 obj=0x40d71628 self=0x22eea8
03-30 12:48:56.644: I/dalvikvm(15359): | sysTid=15404 nice=10 sched=0/0 cgrp=bg_non_interactive handle=2056240
03-30 12:48:56.651: I/dalvikvm(15359): | schedstat=( 3387725830 799926751 876 ) utm=327 stm=10 core=0
03-30 12:48:56.651: I/dalvikvm(15359): at java.io.ByteArrayOutputStream.expand(ByteArrayOutputStream.java:~91)
03-30 12:48:56.651: I/dalvikvm(15359): at java.io.ByteArrayOutputStream.write(ByteArrayOutputStream.java:201)
03-30 12:48:56.651: I/dalvikvm(15359): at android.graphics.YuvImage.nativeCompressToJpeg(Native Method)
03-30 12:48:56.667: I/dalvikvm(15359): at android.graphics.YuvImage.compressToJpeg(YuvImage.java:141)
03-30 12:48:56.667: I/dalvikvm(15359): at com.vfranchi.camerafluid.SavePhotoTask.doInBackground(SavePhotoTask.java:78)
03-30 12:48:56.667: I/dalvikvm(15359): at com.vfranchi.camerafluid.SavePhotoTask.doInBackground(SavePhotoTask.java:1)
03-30 12:48:56.667: I/dalvikvm(15359): at android.os.AsyncTask$2.call(AsyncTask.java:287)
03-30 12:48:56.667: I/dalvikvm(15359): at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:305)
03-30 12:48:56.667: I/dalvikvm(15359): at java.util.concurrent.FutureTask.run(FutureTask.java:137)
03-30 12:48:56.667: I/dalvikvm(15359): at android.os.AsyncTask$SerialExecutor$1.run(AsyncTask.java:230)
03-30 12:48:56.667: I/dalvikvm(15359): at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1076)
03-30 12:48:56.667: I/dalvikvm(15359): at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:569)
03-30 12:48:56.667: I/dalvikvm(15359): at java.lang.Thread.run(Thread.java:856)
03-30 12:48:56.683: W/System.err(15359): java.lang.OutOfMemoryError
03-30 12:48:56.683: W/System.err(15359): at java.io.ByteArrayOutputStream.expand(ByteArrayOutputStream.java:91)
03-30 12:48:56.683: W/System.err(15359): at java.io.ByteArrayOutputStream.write(ByteArrayOutputStream.java:201)
03-30 12:48:56.691: W/System.err(15359): at android.graphics.YuvImage.nativeCompressToJpeg(Native Method)
03-30 12:48:56.691: W/System.err(15359): at android.graphics.YuvImage.compressToJpeg(YuvImage.java:141)
03-30 12:48:56.691: W/System.err(15359): at com.vfranchi.camerafluid.SavePhotoTask.doInBackground(SavePhotoTask.java:78)
03-30 12:48:56.691: W/System.err(15359): at com.vfranchi.camerafluid.SavePhotoTask.doInBackground(SavePhotoTask.java:1)
03-30 12:48:56.691: W/System.err(15359): at android.os.AsyncTask$2.call(AsyncTask.java:287)
03-30 12:48:56.698: W/System.err(15359): at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:305)
03-30 12:48:56.698: W/System.err(15359): at java.util.concurrent.FutureTask.run(FutureTask.java:137)
03-30 12:48:56.698: W/System.err(15359): at android.os.AsyncTask$SerialExecutor$1.run(AsyncTask.java:230)
03-30 12:48:56.698: W/System.err(15359): at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1076)
03-30 12:48:56.698: W/System.err(15359): at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:569)
03-30 12:48:56.706: W/System.err(15359): at java.lang.Thread.run(Thread.java:856)
03-30 12:48:56.706: D/skia(15359): ------- write threw an exception
03-30 12:49:15.181: W/jdwp(15359): Debugger is telling the VM to exit with code=1
03-30 12:49:15.181: I/dalvikvm(15359): GC lifetime allocation: 5 bytes
YuvImage 缓冲区是在 onPreviewFrame 上创建的，如下所示:
public void onPreviewFrame(byte[] data, Camera camera)
{
Camera.Parameters parameters = camera.getParameters();
Camera.Size size = parameters.getPreviewSize();
// Generate a YuvImage from the camera data
int w = size.width;
int h = size.height;
YuvImage photoImage = new YuvImage(data, parameters.getPreviewFormat(), w, h, null);
mPhotoBuffer.add(0, photoImage);
if (mPhotoBuffer.size() > mBufferSize)
{
mPhotoBuffer.remove(mPhotoBuffer.size() - 1);
}
}
最佳答案
我设法为我的问题找到了解决方案，而且是通过不使用 YuvImage 类。我猜 compressToJpeg() 方法有问题，至少在我的设备上我不知道。
这是我的 doInBackground 方法的最终代码:
protected Void doInBackground(Integer... params)
{
int lineHeight = mHeight / mBufferSize;
int currentHeight = 0;
Bitmap result = Bitmap.createBitmap(mWidth, mHeight, Bitmap.Config.ARGB_8888);
Canvas c = new Canvas(result);
Log.d("output", "saving photo: "+mWidth+", "+mHeight);
File outputPath = new File(Environment.getExternalStorageDirectory().getPath() + "/camerafluid-output");
outputPath.mkdirs();
for (int i = 0; i < mBufferSize; i++)
{
File imageFile = new File(mCachePath, "image-"+i);
File outputFile = new File(outputPath, "image-"+i+".jpg");
if (!imageFile.exists())
{
Log.d("output", "image "+i+" not found on cache directory");
continue;
}
try
{
int size = (int)imageFile.length();
byte[] imageBytes = new byte[size];
BufferedInputStream buf = new BufferedInputStream(new FileInputStream(imageFile));
buf.read(imageBytes, 0, size);
buf.close();
int[] imagePixels = MainActivity.convertYUV420_NV21toRGB8888(imageBytes, mWidth, mHeight);
imageBytes = null;
Bitmap bitmapImage = Bitmap.createBitmap(imagePixels, mWidth, mHeight, Bitmap.Config.ARGB_8888);
imagePixels = null;
ByteArrayOutputStream out = new ByteArrayOutputStream();
if (!bitmapImage.compress(CompressFormat.JPEG, 50, out))
{
Log.d("output", "problem converting yuv to jpg");
break;
}
FileOutputStream s = new FileOutputStream(outputFile);
s.write(out.toByteArray());
s.flush();
s.close();
s = null;
currentHeight += lineHeight;
Log.d("output", "currentHeight: "+currentHeight);
publishProgress((int)((i / (float)mBufferSize) * 100));
System.gc();
}
catch(Exception e)
{
}
if (isCancelled())
{
break;
}
}
return null;
}
方法是将 Yuv 数据转换为 RGB 并创建位图，以便我可以压缩为 JPEG 并保存到外部存储器。这仅用于测试目的，这样就不会出现 Signal error 11 或 OutOfMemory 异常。我已经尝试使用 40 张图像的缓冲区。
在此处找到 YUV -> RGB 转换器 Displaying YUV Image in Android .我正在使用 minhaz 想法在处理图像之前在磁盘上缓存缓冲区以避免高内存使用率。
关于android - 一次在多个 YuvImage 上使用 compresstojpeg 时内存不足，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/15696165/