Android 从YV12或NV12字节数组裁剪图像
我实现了Camera.PreviewCallback,然后在其中以字节数组的形式获取原始图像(YV12或NV12格式)。我正在寻找一种方法来裁剪图像的一部分,而不将其转换为位图。图像的裁剪部分将以流形式传输到其他位置(再次作为字节数组) 谢谢你的帮助 公共类CameraAccess实现Camera.PreviewCallback, LoaderCallback接口{Android 从YV12或NV12字节数组裁剪图像,android,image-processing,crop,Android,Image Processing,Crop,我实现了Camera.PreviewCallback,然后在其中以字节数组的形式获取原始图像(YV12或NV12格式)。我正在寻找一种方法来裁剪图像的一部分,而不将其转换为位图。图像的裁剪部分将以流形式传输到其他位置(再次作为字节数组) 谢谢你的帮助 公共类CameraAccess实现Camera.PreviewCallback, LoaderCallback接口{ private byte[] lastFrame; @Override public void onPreviewFrame(b
private byte[] lastFrame;
@Override
public void onPreviewFrame(byte[] frame, Camera arg1) {
synchronized(this) {
this.lastFrame = frame;
}
}
@Override
public byte[] cropFrame(Integer x, Integer y, Integer width, Integer height) {
synchronized(this) {
// how to crop directly from byte array?
}
}
}而图像作为字节数组就是一个巨大数组中图像的每个像素。它从左上角的像素开始,移动到右侧,然后向下移动下一行(回到左侧) 因此,要裁剪它,只需将想要的像素复制到一个新的字节数组中,并使用一些for循环:
Rect cropArea = ... //the are to crop
int currentPos = 0;
byte[] croppedOutput = new byte[cropArea.width() * cropArea.height()];
for(int y = 0; y < height; y++){
for(int x = 0; x < width; x++){
// here you compare if x and y are within the crop area you want
if(cropArea.contains(x, y)){
croppedOutput[currentPos] = frame[positionInArrayForXY(x, y)]
}
}
}
Rect cropera=//庄稼即将收割
int currentPos=0;
byte[]cropedOutput=新字节[cropera.width()*cropera.height()];
对于(int y=0;y
对于方法positioninarayforxy
,您必须做一些额外的数学运算,该方法几乎是x*y
,但当值为零时,必须考虑这些
注:我相信帧是每像素1字节,但不确定,所以如果它是每像素2字节,就有一些额外的数学。但是想法是一样的,你可以从中发展
编辑:
关于你的评论:
不,这东西没有标题,只是像素。这就是为什么它总是给你相机的信息,所以你可以知道大小
当然,这不会影响我的回答,当我回答我希望YUV像RGB一样遵循数组顺序时
我做了一些额外的研究,您可以看到进行YUV到RGB转换的方法,如果仔细检查,您会注意到它每12位使用一次,即1.5字节=>921600*1.5=1382400
基于此,我可以想出一些方法:
- (最容易实现)将您的帧转换为RGB(我知道您指定了您不想要的,但这样会更容易),然后根据我的答案进行裁剪,然后对其进行流式处理
- (最大的开销,一点也不容易)如果流的接收器必须在YUV中接收,则执行上述操作,但在流之前执行链接方法的反转操作,将其转换回YUV
- (非常很难实现,但按照您最初的问题解决)根据我的示例代码、我发布的链接上的代码以及每像素需要12位的事实,使用2个for循环开发代码来进行裁剪
- 有人问我的最终解决方案和一些源代码。这就是我所做的
场景:我的项目基于运行Android的片上系统。我为通过USB连接到主板的本地摄像头实现了摄像头处理。这个摄像头就像安卓智能手机上的摄像头一样工作。第二个是基于IP的摄像头,通过网络传输图像。因此,软件设计可能看起来有点混乱。请随意提问
解决方案:由于OpenCV处理、相机初始化以及颜色和位图转换是一件棘手的事情,我最终将所有内容封装到两个类中,从而避免了在我的Android代码中多次出现愚蠢的代码
第一个类处理颜色/位图和OpenCV矩阵转换。它的定义是:
import org.opencv.android.CameraBridgeViewBase.CvCameraViewFrame;
import org.opencv.core.Mat;
import android.graphics.Bitmap;
public interface CameraFrame extends CvCameraViewFrame {
Bitmap toBitmap();
@Override
Mat rgba();
@Override
Mat gray();
}
所有颜色和位图转换都在该接口的实现范围内。实际转换由OpenCV for Android附带的UTIL完成。您将看到我只使用了一个位图。这是因为节省资源和位图转换是CPU密集型的。所有UI组件都显示/呈现此位图。只有当任何组件请求位图时,才会进行转换
private class CameraAccessFrame implements CameraFrame {
private Mat mYuvFrameData;
private Mat mRgba;
private int mWidth;
private int mHeight;
private Bitmap mCachedBitmap;
private boolean mRgbaConverted;
private boolean mBitmapConverted;
@Override
public Mat gray() {
return mYuvFrameData.submat(0, mHeight, 0, mWidth);
}
@Override
public Mat rgba() {
if (!mRgbaConverted) {
Imgproc.cvtColor(mYuvFrameData, mRgba,
Imgproc.COLOR_YUV2BGR_NV12, 4);
mRgbaConverted = true;
}
return mRgba;
}
// @Override
// public Mat yuv() {
// return mYuvFrameData;
// }
@Override
public synchronized Bitmap toBitmap() {
if (mBitmapConverted)
return mCachedBitmap;
Mat rgba = this.rgba();
Utils.matToBitmap(rgba, mCachedBitmap);
mBitmapConverted = true;
return mCachedBitmap;
}
public CameraAccessFrame(Mat Yuv420sp, int width, int height) {
super();
mWidth = width;
mHeight = height;
mYuvFrameData = Yuv420sp;
mRgba = new Mat();
this.mCachedBitmap = Bitmap.createBitmap(width, height,
Bitmap.Config.ARGB_8888);
}
public synchronized void put(byte[] frame) {
mYuvFrameData.put(0, 0, frame);
invalidate();
}
public void release() {
mRgba.release();
mCachedBitmap.recycle();
}
public void invalidate() {
mRgbaConverted = false;
mBitmapConverted = false;
}
};
摄影机处理封装在两个特殊类中,后面将对此进行解释。一个(HardwareCamera实现ICamera)处理摄像头初始化和关闭,而第二个(CameraAccess)处理OpenCV初始化和其他组件的通知(CameraCanvasView扩展CanvasView实现CameraFrameCallback)有兴趣接收相机图像并在Android视图(UI)中显示它们的用户。此类组件必须连接(注册)到该类
回调(由任何UI组件实现)定义如下:
public interface CameraFrameCallback {
void onCameraInitialized(int frameWidth, int frameHeight);
void onFrameReceived(CameraFrame frame);
void onCameraReleased();
}
此接口的实现由以下UI组件完成:
import android.content.Context;
import android.util.AttributeSet;
import android.view.SurfaceHolder;
import CameraFrameCallback;
public class CameraCanvasView extends CanvasView implements CameraFrameCallback {
private CameraAccess mCamera;
private int cameraWidth = -1;
private int cameraHeight = -1;
private boolean automaticReceive;
private boolean acceptNextFrame;
public CameraCanvasView(Context context, AttributeSet attributeSet) {
super(context, attributeSet);
}
public CameraAccess getCamera() {
return mCamera;
}
public boolean getAcceptNextFrame() {
return acceptNextFrame;
}
public void setAcceptNextFrame(boolean value) {
this.acceptNextFrame = value;
}
public void setCamera(CameraAccess camera, boolean automaticReceive) {
if (camera == null)
throw new NullPointerException("camera");
this.mCamera = camera;
this.mCamera.setAutomaticReceive(automaticReceive);
this.automaticReceive = automaticReceive;
}
@Override
public void onCameraInitialized(int frameWidth, int frameHeight) {
cameraWidth = frameWidth;
cameraHeight = frameHeight;
setCameraBounds();
}
public void setCameraBounds() {
int width = 0;
int height = 0;
if (fixedWidth > 0 && fixedHeight > 0) {
width = fixedWidth;
height = fixedHeight;
} else if (cameraWidth > 0 && cameraHeight > 0) {
width = fixedWidth;
height = fixedHeight;
}
if (width > 0 && height > 0)
super.setCameraBounds(width, height, true);
}
@Override
public void onFrameReceived(CameraFrame frame) {
if (acceptNextFrame || automaticReceive)
super.setBackground(frame);
// reset
acceptNextFrame = false;
}
@Override
public void onCameraReleased() {
setBackgroundImage(null);
}
@Override
public void surfaceCreated(SurfaceHolder arg0) {
super.surfaceCreated(arg0);
if (mCamera != null) {
mCamera.addCallback(this);
if (!automaticReceive)
mCamera.receive(); // we want to get the initial frame
}
}
@Override
public void surfaceDestroyed(SurfaceHolder arg0) {
super.surfaceDestroyed(arg0);
if (mCamera != null)
mCamera.removeCallback(this);
}
}
<?xml version="1.0" encoding="utf-8"?>
<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
android:layout_width="match_parent"
android:layout_height="match_parent"
android:orientation="vertical" >
<eu.level12.graphics.laser.CameraCanvasView
android:id="@+id/my_camera_view"
android:layout_width="match_parent"
android:layout_height="match_parent"
/>
</LinearLayout>
该UI组件可在XML布局中使用,如下所示:
import android.content.Context;
import android.util.AttributeSet;
import android.view.SurfaceHolder;
import CameraFrameCallback;
public class CameraCanvasView extends CanvasView implements CameraFrameCallback {
private CameraAccess mCamera;
private int cameraWidth = -1;
private int cameraHeight = -1;
private boolean automaticReceive;
private boolean acceptNextFrame;
public CameraCanvasView(Context context, AttributeSet attributeSet) {
super(context, attributeSet);
}
public CameraAccess getCamera() {
return mCamera;
}
public boolean getAcceptNextFrame() {
return acceptNextFrame;
}
public void setAcceptNextFrame(boolean value) {
this.acceptNextFrame = value;
}
public void setCamera(CameraAccess camera, boolean automaticReceive) {
if (camera == null)
throw new NullPointerException("camera");
this.mCamera = camera;
this.mCamera.setAutomaticReceive(automaticReceive);
this.automaticReceive = automaticReceive;
}
@Override
public void onCameraInitialized(int frameWidth, int frameHeight) {
cameraWidth = frameWidth;
cameraHeight = frameHeight;
setCameraBounds();
}
public void setCameraBounds() {
int width = 0;
int height = 0;
if (fixedWidth > 0 && fixedHeight > 0) {
width = fixedWidth;
height = fixedHeight;
} else if (cameraWidth > 0 && cameraHeight > 0) {
width = fixedWidth;
height = fixedHeight;
}
if (width > 0 && height > 0)
super.setCameraBounds(width, height, true);
}
@Override
public void onFrameReceived(CameraFrame frame) {
if (acceptNextFrame || automaticReceive)
super.setBackground(frame);
// reset
acceptNextFrame = false;
}
@Override
public void onCameraReleased() {
setBackgroundImage(null);
}
@Override
public void surfaceCreated(SurfaceHolder arg0) {
super.surfaceCreated(arg0);
if (mCamera != null) {
mCamera.addCallback(this);
if (!automaticReceive)
mCamera.receive(); // we want to get the initial frame
}
}
@Override
public void surfaceDestroyed(SurfaceHolder arg0) {
super.surfaceDestroyed(arg0);
if (mCamera != null)
mCamera.removeCallback(this);
}
}
<?xml version="1.0" encoding="utf-8"?>
<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
android:layout_width="match_parent"
android:layout_height="match_parent"
android:orientation="vertical" >
<eu.level12.graphics.laser.CameraCanvasView
android:id="@+id/my_camera_view"
android:layout_width="match_parent"
android:layout_height="match_parent"
/>
</LinearLayout>
备注:我知道这不是一个完整的实现,但我希望你能理解这一点。最有趣的部分是颜色转换,这可以在这篇文章的顶部找到。但是通常标题应该包含一些关于图像格式的信息,或者?无论如何,我的预览是1280x720,这将导致921600字节的数组长度。但预览帧的长度为1382399字节。这不适合你的解决方案,对吗?谢谢你的高级研究。好消息。我将尝试直接从YUV字节数组中裁剪。我想避免发送端的RGB转换,因为它会占用太多的CPU。工作正常,但我想知道它的性能。因为它会查看原始图像的每个像素,但是如果for循环的开始和限制计算正确,那么for循环和for循环都可以在裁剪区域中迭代。我完全同意你的观点。如果您已经知道如何将XY位置映射到YUV阵列,那么下一个优化步骤就是仅在该区域上迭代。广告使代码复杂化,但可以肯定的是
@Override
protected void onResume() {
super.onResume();
if (fourPointView != null) {
cameraAccess = CameraAccess.getInstance(this);
canvasView.setCamera(cameraAccess, true);
} else {
cameraAccess = null;
}
if (cameraAccess != null)
cameraAccess.setAutomaticReceive(true);
if (cameraAccess != null && fourPointView != null)
cameraAccess.setRegionOfInterest(RectTools.toRect(canvasView
.getCamera().getViewport()));
}
@Override
protected void onPause() {
super.onPause();
if (cameraAccess != null)
cameraAccess.setRegionOfInterest(null);
}