说明
因工作方面接触到图像处理这一块,需要对手机摄像头采集的原始帧做Rotate或者Scale,但无奈对此的了解少之又少,于是网上搜了一顿,完事后将最近所学总结一下,以方便之后的人别踩太多坑。
首先想要了解YUV为何物:请猛戳我
上面的链接中微软已经写的很详细了,国内大部分文章都是翻译这篇文章的,如果还有疑问的同学可以参考下面这些大神的博客:
看完上面的文章应该都会有所了解和认识了,因为在Android SDK <= 20(Android5.0)中Google支持的Camera Preview Callback的YUV常用格式有两种:NV21 / YV12,在此针对这两种格式做分析。
NV21:
引用一段微软的叙述:
4:2:0 Formats, 12 Bits per Pixel
Four 4:2:0 12-bpp formats are recommended, with the following FOURCC codes:
- IMC2
- IMC4
- YV12
- NV12
In all of these formats, the chroma channels are subsampled by a factor of two in both the horizontal and vertical dimensions.
YV12
All of the Y samples appear first in memory as an array of unsigned char values.
This array is followed immediately by all of the V (Cr) samples. The stride of the V plane is half the stride of the Y plane, and the V plane contains half as many lines as the Y plane. The V plane is followed immediately by all of the U (Cb) samples, with the same stride and number of lines as the V plane (Figure 12).
Figure 12:
NV12
All of the Y samples are found first in memory as an array of unsigned char values with an even number of lines.
The Y plane is followed immediately by an array of unsigned char values that contains packed U (Cb) and V (Cr) samples, as shown in Figure 13. When the combined U-V array is addressed as an array of little-endian WORD values, the LSBs contain the U values, and the MSBs contain the V values. NV12 is the preferred 4:2:0 pixel format for DirectX VA. It is expected to be an intermediate-term requirement for DirectX VA accelerators supporting 4:2:0 video.
Figure 13:
从上可知YV12和NV12所占内存是12bits/Pixel,因为每个Y就是一个像素点,注意着色加粗的叙述,YUV值在内存中是按照数组的形式存放的,而由于YV12和NV21都是属于planar格式,也就是Y值和UV值是独立采样的:
In a planar format, the Y, U, and V components are stored as three separate planes.
既然Y、U、V值都是独立的,那就意味着我们可以分别处理相应的值,比如在YV12中,排列方式是这样的,每4个Y共用一对UV值,而U、V值又是按照如下格式排列
下面是YV12格式中宽为16,高为4像素的排列
Y第一行: | Y Y | Y Y | Y Y | Y Y |
---|---|---|---|---|
Y第二行: | Y Y | Y Y | Y Y | Y Y |
Y第三行: | Y Y | Y Y | Y Y | Y Y |
Y第四行: | Y Y | Y Y | Y Y | Y Y |
V第一行: | V0 | V1 | V2 | V3 |
U第一行: | U0 | U1 | U2 | U3 |
V第二行: | V4 | V5 | V6 | V7 |
U第二行: | U4 | U5 | U6 | U7 |
既然知道了YUV值的结构,我们就可以任性的对此图像做Rotate,scale等等。这里我以480x270 (16:9)的一张原始帧图像举例,贴出部分代码示例:
任意设定的一个带有onPreviewFrame的类,CameraPreviewFrame.java:1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76/**
* 获取preview的原始帧:
*
* 这里有个前提,因为Android camera preview默认格式为NV21的,所以需要
* 调用setPreviewFormat()方法设置为我们需要的格式
*
*/
public void onPreviewFrame(byte[] data, Camera camera) {// 假设这里的data为480x270原始帧
String SRC_FRAME_WIDTH = 480;
String SRC_FRAME_HEIGHT = 270;
String DES_FRAME_WIDTH = 480;
String DES_FRAME_HEIGHT = 270;
// 此处将data数组保存在了指定的路径,保存类型为jpeg格式,但是普通的图片浏
// 览器是无法打开的,需要使用RawView等专业的工具打开。
saveImageData(data);
// 定义与原始帧大小一样的outputData,因为YUV420所占内存是12Bits/Pixel,
// 每个Y为一个像素8bit=1Byte,U=2bit=1/4(Byte),V=2bit=1/4(Byte),
// Y值数量为480*270,则U=V=480*270*(1/4)
byte[] outputData = new byte[DES_FRAME_WIDTH * DES_FRAME_HEIGHT * 3 / 2];
// call the JNI method to rotate frame data clockwise 90 degrees
YuvUtil.DealYV12(data, outputData, SRC_FRAME_WIDTH, SRC_FRAME_HEIGHT, 90);
saveImageData(outputData);
}
}
// save image to sdcard path: Pictures/MyTestImage/
public void saveImageData(byte[] imageData) {
File imageFile = getOutputMediaFile(MEDIA_TYPE_IMAGE);
if (imageFile == null) {
return;
}
try {
FileOutputStream fos = new FileOutputStream(imageFile);
fos.write(imageData);
fos.close();
} catch (FileNotFoundException e) {
e.printStackTrace();
Log.e(TAG, "File not found: " + e.getMessage());
} catch (IOException e) {
e.printStackTrace();
Log.e(TAG, "Error accessing file: " + e.getMessage());
}
}
public static File getOutputMediaFile(int type) {
File imageFileDir = new File(Environment.getExternalStoragePublicDirectory(Environment.DIRECTORY_PICTURES), "MyTestImage");
if (!imageFileDir.exists()) {
if (!imageFileDir.mkdirs()) {
Log.e(TAG, "can't makedir for imagefile");
return null;
}
}
// Create a media file name
String timeStamp = new SimpleDateFormat("yyyyMMdd_HHmmss").format(new Date());
File imageFile;
if (type == MEDIA_TYPE_IMAGE) {
imageFile = new File(imageFileDir.getPath() + File.separator +
"IMG_" + timeStamp + ".jpg");
} else if (type == MEDIA_TYPE_VIDEO) {
imageFile = new File(imageFileDir.getPath() + File.separator +
"VID_" + timeStamp + ".mp4");
} else {
return null;
}
return imageFile;
}
上面的代码中可以看到调用了JNI的方法:YuvUtil.RotateYV12()
1 |
|
com_example_jni_YuvUtil.h1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
/* DO NOT EDIT THIS FILE - it is machine generated */
/* Header for class _Included_com_example_jni_YuvUtil */
extern "C" {
/*
* Class: com_example_jni_YuvUtil
* Method: initYV12
* Signature: (II)V
*/
JNIEXPORT void JNICALL Java_com_example_jni_YuvUtil_initYV12
(JNIEnv *, jclass, jint, jint);
/*
* Class: com_example_jni_YuvUtil
* Method: DealYV12
* Signature: ([B[BIIIII)V
*/
JNIEXPORT void JNICALL Java_com_example_jni_YuvUtil_DealYV12
(JNIEnv *, jclass, jbyteArray, jbyteArray, jint, jint, jint, jint, jint);
}
```
**com_example_jni_YuvUtil.c**
``` c
char *input_src_data, *output_src_data, *src_y_data,
*src_u_data, *src_v_data, *dst_y_data, *dst_v_data;
int src_data_width, src_data_height, len_src;
/*
* Class: com_example_jni_YuvUtil
*/
JNIEXPORT void JNICALL Java_com_example_jni_YuvUtil_initYV12
(JNIEnv *env, jclass jcls, jint length, jint scaleDataLength) {
len_src = length;
len_scale = scaleDataLength;
LOGD("########## len_src = %d, len_scale = %d \n", len_src, len_scale);
input_src_data = malloc(sizeof(char) * len_src);
LOGD("########## input_src_data = %d \n", input_src_data);
src_y_data = malloc(sizeof(char) * (len_src * 2 / 3));
src_u_data = malloc(sizeof(char) * (len_src / 6));
src_v_data = malloc(sizeof(char) * (len_src / 6));
dst_y_data = malloc(sizeof(char) * (len_src * 2 / 3));
dst_u_data = malloc(sizeof(char) * (len_src / 6));
dst_v_data = malloc(sizeof(char) * (len_src / 6));
}
JNIEXPORT void JNICALL Java_com_example_jni_YuvUtil_DealYV12
(JNIEnv *env, jclass jcls, jbyteArray src_data,
jbyteArray dst_data, jint width, jint height, jint rotation, jint dst_width, jint dst_height) {
src_data_width = width;
src_data_height = height;
// 将src_data的数据传给input_src_data
(*env)->GetByteArrayRegion (env, src_data, 0, len_src, (jbyte*)(input_src_data));
/*以下三个memcpy分别将Y、U、V值从src_data中提取出来,将YUV值分别scale或者rotate,则可得到对应格式的图像数据*/
// get y plane
memcpy(src_y_data, input_src_data , (len_src * 2 /3));
// get u plane
memcpy(src_u_data, input_src_data + (len_src * 2 / 3), len_src / 6);
// get v plane
memcpy(src_v_data, input_src_data + (len_src * 5 / 6 ), len_src / 6);
/*获取yuv三个值的数据可以做相应操作*/
// .........
// .........
// 例:将Y值置为0,则得到没有灰度的图像;
memset(input_src_data + src_data_width * src_data_height, 0, src_data_width * src_data_height);
// 将input_src_data的数据返回给dst_data输出
// output to the dst_data
(*env)->SetByteArrayRegion (env, dst_data, 0, len_src, (jbyte*)(input_src_data));
}
/**
* free memory
*/
JNIEXPORT void JNICALL Java_com_example_jni_YuvUtil_ReleaseYV12
(JNIEnv *env , jclass jcls) {
free(output_src_data);
free(input_src_data);
}
注意:以上代码不是完全的,只是用于说明而已,如果需要更多的操作还请各位朋友自己完善,因为没怎么写过这类博客代码很乱,如有表述的不清楚和有问题的地方,大家可以给我留言。