问题来源：
在进行Android camera相关的开发时，对于图片数据不论是缓存在本地磁盘还是上传到后端，都需要先对图片进行压缩处理。但是JPG（JPEG）图片在压缩后原图的EXIF信息也会丢失。那如果想保留exif数据该怎么处理？
关键词描述
EXIF：可交换图像文件格式（英语：Exchangeable image file format，官方简称Exif），是专门为数码相机的照片设定的，可以附加于JPEG、TIFF、RIFF等文件之中，可以记录数码照片的属性信息和拍摄数据。比如记录以下信息：
| 项目 | 资讯（举例） |
|---|---|
| 制造厂商 | Canon |
| 相机型号 | Canon EOS-1Ds Mark III |
| 影像方向 | 正常（upper-left） |
| 影像解析度X | 300 |
| 影像解析度Y | 300 |
| 解析度单位 | dpi |
| 软件 | Adobe Photoshop CS Macintosh |
| 最后异动时间 | 2005:10:06 12:53:19 |
| YCbCrPositioning | 2 |
| 曝光时间 | 0.00800 (1/125) sec |
| 光圈值 | F22 |
| 拍摄模式 | 光圈优先 |
| ISO感光值 | 100 |
| Exif资讯版本 | 30,32,32,31 |
| 影像拍摄时间 | 2005:09:25 15:00:18 |
| 影像存入时间 | 2005:09:25 15:00:18 |
| 曝光补偿（EV+-） | 0 |
| 测光模式 | 点测光（Spot） |
| 闪光灯 | 关闭 |
| 镜头实体焦长 | 12 mm |
| Flashpix版本 | 30,31,30,30 |
| 影像色域空间 | sRGB |
| 影像尺寸X | 5616 pixel |
| 影像尺寸Y | 3744 pixel |
现已有方案
利用Google提供的 android.support.media.ExifInterface 对图片的exif进行读写设置
This is a class for reading and writing Exif tags in a JPEG file or a RAW image file.
Supported formats are: JPEG, DNG, CR2, NEF, NRW, ARW, RW2, ORF, PEF, SRW and RAF.
Attribute mutation is supported for JPEG image files.
但是这个封装类只提供了 getXXX() 和 setAttributes(String tag, String value) 这种操作单个属性的方法，如果想将原图片文件中的所有exif信息完整复制到另一个图片中会非常繁琐。因此有人通过反射，对所有属性名进行遍历，从而实现了批量操作。也算是一种解决方案，具体如下：
public static void saveExif(String oldFilePath, String newFilePath) throws Exception {
ExifInterface oldExif = new ExifInterface(oldFilePath);
ExifInterface newExif = new ExifInterface(newFilePath);
Class cls = ExifInterface.class;
Field[] fields = cls.getFields();
for (int i = 0; i < fields.length; i++) {
String fieldName = fields[i].getName();
if (!TextUtils.isEmpty(fieldName) && fieldName.startsWith("TAG")) {
String fieldValue = fields[i].get(cls).toString();
String attribute = oldExif.getAttribute(fieldValue);
if (attribute != null) {
newExif.setAttribute(fieldValue, attribute);
}
}
}
//将内存中的修改写入磁盘（IO操作）
newExif.saveAttributes();
}
但是以上方案弊端也很明显，就是需要对文件进行多次IO操作。为什么这么说？
首先观察上面方法中的两个参数都是文件路径，意思就是我们在拍完照通过 onPictureTaken(byte[] data, Camera camera) 回调方法拿到图片的 byte[] data 数据后的workflow是这样的：
- 将data缓存到磁盘，路径为
oldFilePath；（IO） - 将data转换成
bitmap进行压缩、旋转、剪切等操作； - 将处理后的
bitmap缓存到磁盘，路径为newFilePath；(IO) - 调用上面的
saveExif(oldFilePath, newFilePath)方法; (IO)
能否只在内存中操作？发现有 ExifInterface (String filename) 和 ExifInterface (InputStream inputStream) 两种构造方法, 所以我尝试进行如下改造：
public static void saveExif(byte[] srcData, byte[] destData) throws Exception {
ExifInterface oldExif = new ExifInterface(new ByteArrayInputStream(srcData));
ExifInterface newExif = new ExifInterface(new ByteArrayInputStream(destData));
...
newExif.saveAttributes();
}
然鹅并没有什么卵用， 直接抛异常，后研究源码发现 saveAttributes() 的流程是这样的：
- 校验构造方法中传入的
fileName是否为空，若为空则抛异常；假设我们new ExifInterface (“/a/b/picture.jpg”)，即fileName为/a/b/picture.jpg,; - 把
/a/b/picture.jpg重命名为/a/b/picture.jpg.tmp； - 新建
/a/b/picture.jpg文件； - 将
/a/b/picture.jpg.tmp文件中的数据加上修改后的exif 存入到新建的/a/b/picture.jpg文件中； - 删除
/a/b/picture.jpg.tmp；
由此可见， saveAttributes() 必然是IO操作，而且对于EXIF的修改只能使用第一种构造方式，即必须传入文件路径. 否则必然抛出异常。所以进一步改造如下：
public static void saveExif(byte[] srcData, String destFilePath) throws Exception {
ExifInterface oldExif = new ExifInterface(new ByteArrayInputStream(srcData));
ExifInterface newExif = new ExifInterface(destFilePath);
...
newExif.saveAttributes();
}
结果可行，而且少了一次IO （第一步）; 但是我觉得还不够优雅。。。
我的解决方案
我的目标是将所有有关图片的操作都放到内存中完成，最后只缓存一份图片数据。
思路很简单，不管是图片还是其他文件，其本质都是格式化的数据，都有其专用的数据结构。那么就去研究下JPG的数据结构好了，只要找到 exif 数据块的起始索引，然后从源文件byte[]中复制插入到目标文件byte[]对应位置中不就ok了。
如上图所示，每一个JPEG文件的内容都开始于一个二进制的值 '0xFFD8', 并结束与二进制值'0xFFD9'. 在JPEG的数据 中有好几种类似于二进制 0xFFXX 的数据, 它们都统称作 "标记", 并且它们代表了一段JPEG的 信息数据. 0xFFD8 的意思是 SOI图像起始(Start of image), 0xFFD9 则表示 EOI图像结束 (End of image). 这两个特殊的标记的后面都不跟随数据, 而其他的标记在后面则会附带数据. 标记的基本格式如下.
0xFF+标记号(1个字节)+数据大小描述符(2个字节)+数据内容(n个字节)
而对于EXIF数据，使用的是APP1标记，前两个字节固定为 0xFFE1，后面紧跟着两个字节记录的是exif数据内容的 length + 2，假设这两个字节的值是 24，那么exif数据内容的长度就是22字节.
了解了JPG的数据格式后，剩下的就是动手操作数组了，找到EXIF在数组中的起始索引，把它抠出来插入到新数组中去！
/**
* 将原图片中的EXIF复制到目标图片中
* 仅限JPEG
* @param srcData
* @param destData
* @return
*/
public static byte[] cloneExif(byte[] srcData, byte[] destData) {
if (srcData == null || srcData.length == 0 || destData == null || destData.length == 0) return null;
ImageHeaderParser srcImageHeaderParser = new ImageHeaderParser(srcData);
byte[] srcExifBlock = srcImageHeaderParser.getExifBlock();
if (srcExifBlock == null || srcExifBlock.length <= 4) return null;
LOG.d(TAG, "pictureData src: %1$s KB; dest: %2$s KB", srcData.length / 1024, destData.length / 1024);
LOG.d(TAG, "srcExif: %s B", srcExifBlock.length);
ImageHeaderParser destImageHeaderParser = new ImageHeaderParser(destData);
byte[] destExifBlock = destImageHeaderParser.getExifBlock();
if (destExifBlock != null && destExifBlock.length > 0) {
LOG.d(TAG, "destExif: %s B", destExifBlock.length);
//目标图片中已有exif信息, 需要先删除
int exifStartIndex = destImageHeaderParser.getExifStartIndex();
//构建新数组
byte[] newDestData = new byte[srcExifBlock.length + destData.length - destExifBlock.length];
//copy 1st block
System.arraycopy(destData, 0, newDestData, 0, exifStartIndex);
//copy 2rd block (exif)
System.arraycopy(srcExifBlock, 0, newDestData, exifStartIndex, srcExifBlock.length);
//copy 3th block
int srcPos = exifStartIndex + destExifBlock.length;
int destPos = exifStartIndex + srcExifBlock.length;
System.arraycopy(destData, srcPos, newDestData, destPos, destData.length - srcPos);
LOG.d(TAG, "output image Data with exif: %s KB", newDestData.length / 1024);
return newDestData;
} else {
LOG.d(TAG, "destExif: %s B", 0);
//目标图片中没有exif信息
byte[] newDestData = new byte[srcExifBlock.length + destData.length];
//copy 1st block (前两个字节)
System.arraycopy(destData, 0, newDestData, 0, 2);
//copy 2rd block (exif)
System.arraycopy(srcExifBlock, 0, newDestData, 2, srcExifBlock.length);
//copy 3th block
int srcPos = 2;
int destPos = 2 + srcExifBlock.length;
System.arraycopy(destData, srcPos, newDestData, destPos, destData.length - srcPos);
LOG.d(TAG, "output image Data with exif: %s KB", newDestData.length / 1024);
return newDestData;
}
}
如此，拿到图片的 byte[] srcData 数据后，整个workflow就简化成：
- 将srcData转换成
bitmap进行压缩、旋转、剪切等操作，后再转成 byte[] destData； - 调用上面的
cloneExif(srcData, destData)方法，将原图的exif复制到压缩处理后的图片中; - 将压缩处理后的含有exif的图片data 缓存到磁盘；（IO）
只进行一次IO操作~
附：ImageHeaderParser 全部代码实现（参考Glide库）
/**
* Created by Jessewo on 2018/11/29.
*
* A class for parsing the exif orientation and other data from an image header.
* Exif文件格式描述
*
* --------------------------------------------------------------------------------------------------------------------------
* | SOI 标记 | 标记 XX 的大小=SSSS | 标记 YY 的大小=TTTT | SOS 标记 的大小=UUUU | 图像数据流 | EOI 标记
* --------------------------------------------------------------------------------------------------------------------------
* | FFD8 | FFXX SSSS DDDD...... | FFYY TTTT DDDD...... | FFDA UUUU DDDD.... | I I I I.... | FFD9
* --------------------------------------------------------------------------------------------------------------------------
*/
public class ImageHeaderParser {
private static final String TAG = "ImageHeaderParser";
private int magicNumber;
/**
* The format of the image data including whether or not the image may include transparent pixels.
*/
public enum ImageType {
/**
* GIF type.
*/
GIF(true),
/**
* JPG type.
*/
JPEG(false),
/**
* PNG type with alpha.
*/
PNG_A(true),
/**
* PNG type without alpha.
*/
PNG(false),
/**
* Unrecognized type.
*/
UNKNOWN(false);
private final boolean hasAlpha;
ImageType(boolean hasAlpha) {
this.hasAlpha = hasAlpha;
}
public boolean hasAlpha() {
return hasAlpha;
}
}
private static final int GIF_HEADER = 0x474946;
private static final int PNG_HEADER = 0x89504E47;
//JPEG 起始标记字节
private static final int EXIF_MAGIC_NUMBER = 0xFFD8;
// "MM".
private static final int MOTOROLA_TIFF_MAGIC_NUMBER = 0x4D4D;
// "II".
private static final int INTEL_TIFF_MAGIC_NUMBER = 0x4949;
// EXIF数据内容的第一部分
private static final String JPEG_EXIF_SEGMENT_PREAMBLE = "Exif\0\0";
private static final byte[] JPEG_EXIF_SEGMENT_PREAMBLE_BYTES;
//JPEG SOS数据流的起始(Start of stream) 标记
private static final int SEGMENT_SOS = 0xDA;
//JPEG 图像结束标记字节
private static final int MARKER_EOI = 0xD9;
private static final int SEGMENT_START_ID = 0xFF;
//EXIF用的APP1标记位, 其后两个字节记录 EXIF 的大小(需要-2)
private static final int EXIF_SEGMENT_TYPE = 0xE1;
private static final int ORIENTATION_TAG_TYPE = 0x0112;
private static final int[] BYTES_PER_FORMAT = {0, 1, 1, 2, 4, 8, 1, 1, 2, 4, 8, 4, 8};
private final StreamReader streamReader;
private ImageType imageType;
private byte[] exifBlock;
private int exifStartIndex;
static {
byte[] bytes = new byte[0];
try {
bytes = JPEG_EXIF_SEGMENT_PREAMBLE.getBytes("UTF-8");
} catch (UnsupportedEncodingException e) {
// Ignore.
}
JPEG_EXIF_SEGMENT_PREAMBLE_BYTES = bytes;
}
public ImageHeaderParser(byte[] data) {
this(new ByteArrayInputStream(data));
}
public ImageHeaderParser(InputStream is) {
streamReader = new StreamReader(is);
parse();
}
private void parse() {
try {
final int magicNumber = streamReader.getUInt16();
this.magicNumber = magicNumber;
ImageType imageType;
// JPEG.
if (magicNumber == EXIF_MAGIC_NUMBER) {
imageType = JPEG;
parseExifBlock();
} else {
final int firstFourBytes = magicNumber << 16 & 0xFFFF0000 | streamReader.getUInt16() & 0xFFFF;
// PNG.
if (firstFourBytes == PNG_HEADER) {
// See: http://stackoverflow.com/questions/2057923/how-to-check-a-png-for-grayscale-alpha-color-type
streamReader.skip(25 - 4);
int alpha = streamReader.getByte();
// A RGB indexed PNG can also have transparency. Better safe than sorry!
imageType = alpha >= 3 ? PNG_A : PNG;
} else if (firstFourBytes >> 8 == GIF_HEADER) {
// GIF from first 3 bytes.
imageType = GIF;
} else {
imageType = UNKNOWN;
}
}
this.imageType = imageType;
} catch (IOException e) {
e.printStackTrace();
}
}
// 0xD0A3C68 -> 4 ? Arrays.copyOfRange(this.exifBlock, 4, this.exifBlock.length - 1) : null;
}
public int getExifStartIndex() {
return this.exifStartIndex;
}
/**
* 将原图片中的EXIF复制到目标图片中
* 仅限JPEG
* @param srcData
* @param destData
* @return
*/
public static byte[] cloneExif(byte[] srcData, byte[] destData) {
if (srcData == null || srcData.length == 0 || destData == null || destData.length == 0) return null;
ImageHeaderParser srcImageHeaderParser = new ImageHeaderParser(srcData);
byte[] srcExifBlock = srcImageHeaderParser.getExifBlock();
if (srcExifBlock == null || srcExifBlock.length <= 4) return null;
LOG.d(TAG, "pictureData src: %1$s KB; dest: %2$s KB", srcData.length / 1024, destData.length / 1024);
LOG.d(TAG, "srcExif: %s B", srcExifBlock.length);
ImageHeaderParser destImageHeaderParser = new ImageHeaderParser(destData);
byte[] destExifBlock = destImageHeaderParser.getExifBlock();
if (destExifBlock != null && destExifBlock.length > 0) {
LOG.d(TAG, "destExif: %s B", destExifBlock.length);
//目标图片中已有exif信息, 需要先删除
int exifStartIndex = destImageHeaderParser.getExifStartIndex();
//构建新数组
byte[] newDestData = new byte[srcExifBlock.length + destData.length - destExifBlock.length];
//copy 1st block
System.arraycopy(destData, 0, newDestData, 0, exifStartIndex);
//copy 2rd block (exif)
System.arraycopy(srcExifBlock, 0, newDestData, exifStartIndex, srcExifBlock.length);
//copy 3th block
int srcPos = exifStartIndex + destExifBlock.length;
int destPos = exifStartIndex + srcExifBlock.length;
System.arraycopy(destData, srcPos, newDestData, destPos, destData.length - srcPos);
LOG.d(TAG, "output image Data with exif: %s KB", newDestData.length / 1024);
return newDestData;
} else {
LOG.d(TAG, "destExif: %s B", 0);
//目标图片中没有exif信息
byte[] newDestData = new byte[srcExifBlock.length + destData.length];
//copy 1st block (前两个字节)
System.arraycopy(destData, 0, newDestData, 0, 2);
//copy 2rd block (exif)
System.arraycopy(srcExifBlock, 0, newDestData, 2, srcExifBlock.length);
//copy 3th block
int srcPos = 2;
int destPos = 2 + srcExifBlock.length;
System.arraycopy(destData, srcPos, newDestData, destPos, destData.length - srcPos);
LOG.d(TAG, "output image Data with exif: %s KB", newDestData.length / 1024);
return newDestData;
}
}
private void parseExifBlock() throws IOException {
short segmentId, segmentType;
int segmentLength;
int index = 2;
while (true) {
segmentId = streamReader.getUInt8();
if (segmentId != SEGMENT_START_ID) {
if (Log.isLoggable(TAG, Log.DEBUG)) {
Log.d(TAG, "Unknown segmentId=" + segmentId);
}
return;
}
segmentType = streamReader.getUInt8();
if (segmentType == SEGMENT_SOS) {
return;
} else if (segmentType == MARKER_EOI) {
if (Log.isLoggable(TAG, Log.DEBUG)) {
Log.d(TAG, "Found MARKER_EOI in exif segment");
}
return;
}
// Segment length includes bytes for segment length.
segmentLength = streamReader.getUInt16() - 2;
if (segmentType != EXIF_SEGMENT_TYPE) {
//跳过所有的非exif标记块
long skipped = streamReader.skip(segmentLength);
if (skipped != segmentLength) {
if (Log.isLoggable(TAG, Log.DEBUG)) {
Log.d(TAG, "Unable to skip enough data"
+ ", type: " + segmentType
+ ", wanted to skip: " + segmentLength
+ ", but actually skipped: " + skipped);
}
return;
}
index += (4 + segmentLength);
} else {
//找到exif block
byte[] segmentData = new byte[segmentLength];
int read = streamReader.read(segmentData);
byte[] block = new byte[2 + 2 + read];
block[0] = (byte) SEGMENT_START_ID;
block[1] = (byte) EXIF_SEGMENT_TYPE;
int length = read + 2;
block[2] = (byte) ((length >> 8) & 0xFF);
block[3] = (byte) (length & 0xFF);
if (read != segmentLength) {
if (Log.isLoggable(TAG, Log.DEBUG)) {
Log.d(TAG, "Unable to read segment data"
+ ", type: " + segmentType
+ ", length: " + segmentLength
+ ", actually read: " + read);
}
} else {
System.arraycopy(segmentData, 0, block, 4, read);
}
this.exifBlock = block;
this.exifStartIndex = index;
return;
}
}
}
/**
* Parse the orientation from the image header. If it doesn't handle this image type (or this is not an image)
* it will return a default value rather than throwing an exception.
*
* @return The exif orientation if present or -1 if the header couldn't be parsed or doesn't contain an orientation
* @throws IOException
*/
public int getOrientation() {
if (!handles(this.magicNumber)) {
return -1;
} else {
byte[] exifData = getExifContent();
boolean hasJpegExifPreamble = exifData != null
&& exifData.length > JPEG_EXIF_SEGMENT_PREAMBLE_BYTES.length;
if (hasJpegExifPreamble) {
for (int i = 0; i < JPEG_EXIF_SEGMENT_PREAMBLE_BYTES.length; i++) {
if (exifData[i] != JPEG_EXIF_SEGMENT_PREAMBLE_BYTES[i]) {
hasJpegExifPreamble = false;
break;
}
}
}
if (hasJpegExifPreamble) {
return parseExifSegment(new RandomAccessReader(exifData));
} else {
return -1;
}
}
}
private static int calcTagOffset(int ifdOffset, int tagIndex) {
return ifdOffset + 2 + 12 * tagIndex;
}
private static boolean handles(int imageMagicNumber) {
return (imageMagicNumber & EXIF_MAGIC_NUMBER) == EXIF_MAGIC_NUMBER
|| imageMagicNumber == MOTOROLA_TIFF_MAGIC_NUMBER
|| imageMagicNumber == INTEL_TIFF_MAGIC_NUMBER;
}
/**
* InputStream的包装类
*
*/
private static class StreamReader {
private final InputStream is;
//motorola / big endian byte order
public StreamReader(InputStream is) {
this.is = is;
}
/**
* 读两个字节, 并将两个byte转化为 int
*
* @return
* @throws IOException
*/
public int getUInt16() throws IOException {
return (is.read() << 8 & 0xFF00) | (is.read() & 0xFF);
}
/**
* 读一个字节，并将一个byte转化为short
*
* @return
* @throws IOException
*/
public short getUInt8() throws IOException {
return (short) (is.read() & 0xFF);
}
public long skip(long total) throws IOException {
if (total < 0) {
return 0;
}
long toSkip = total;
while (toSkip > 0) {
long skipped = is.skip(toSkip);
if (skipped > 0) {
toSkip -= skipped;
} else {
// Skip has no specific contract as to what happens when you reach the end of
// the stream. To differentiate between temporarily not having more data and
// having finished the stream, we read a single byte when we fail to skip any
// amount of data.
int testEofByte = is.read();
if (testEofByte == -1) {
break;
} else {
toSkip--;
}
}
}
return total - toSkip;
}
public int read(byte[] buffer) throws IOException {
int toRead = buffer.length;
int read;
while (toRead > 0 && ((read = is.read(buffer, buffer.length - toRead, toRead)) != -1)) {
toRead -= read;
}
return buffer.length - toRead;
}
public int getByte() throws IOException {
return is.read();
}
}
}
参考：
- 维基百科：EXIF
- Exif文件格式描述
- glide
- android.support.media.ExifInterface
- Android Bitmap小技巧 - 压缩时保留图片的Exif信息