bayer转dng实现过程记录
前言
项目中需要将imx185出来的raw数据转成dng格式,一开始认为很简单的事情,后面才发现还是挺复杂的!!!首先考虑的是不写任何代码,直接用adobe提供的转换工具来转,结果发现,不仅是adobe的工具,网上能够找到的工具(所谓的raw转dng)都是针对各大厂商的raw,而不能支持最原始的bayer raw。于是又考虑了两种解决方式,第一种就是将bayer raw随便转换成某一个厂商的raw(CR2, CRW, DNG, MRW, NEF, ORF, RAF, DCR, ARW中的一种),然后用所谓的raw转dng工具再转dng,结果发现厂商的raw一般都是不开放的,网上虽然有一些逆向工程,但不完整,没办法,只好用第二种方案了。第二种方案是考虑基于adobe提供的dng_sdk开发一个bayer转dng的程序,但是这个工作量真的很大,首先得完全弄懂dng格式,其次,adobe的sdk没有任何文档(html文件夹下有点点,但那个没多大作用),网上也找到不任何关于该sdk的说明!没办法,只好边看dng_spec_1.4.0.0边一步步解决。幸运的时,偶然在论坛上发现了两个关于bayer转dng的讨论,根据他们的代码,实现了转换。
dng_sdk linux移植具体步骤
adobe官方提供的dng_sdk只有windows和mac版本,而我用的是linux系统,所以得想办法将其移植到linux里去。网上有有两种方案可以参考,方案一,一步一步按照他说的移植就ok了,方案二,也是一步一步根据他里面的文档并结合dng_sdk文档来移植。两种方案我都有试过,都是可行的,下面我主要基于第二种方案来说说具体的实践过程。
安装依赖jpeg
在ubuntu下,要安装jpeg依赖很简单,一条命令搞定
sudo apt-get install cmake libjpeg8-dev uuid-dev
安装依赖xmp sdk
下载xmp_sdk
wget http://download.macromedia.com/pub/developer/xmp/sdk/XMP-Toolkit-SDK-CC-201306.zip
unzip XMP-Toolkit-SDK-CC-201306.zip
mv XMP-Toolkit-SDK-CC201306 xmp_sdk
然后根据xmp_sdk/third-party下各子目录里的ReadMe.txt安装其他一些依赖
zlib
cd xmp_sdk/third-party/zlib
wget http://zlib.net/zlib-1.2.8.tar.gz
tar xzf zlib-1.2.8.tar.gz
cp zlib-1.2.8/*.h zlib-1.2.8/*.c .
expat
cd xmp_sdk/third-party/expat
Download here the source at http://sourceforge.net/projects/expat/files/expat/2.1.0/expat-2.1.0.tar.gz/download.
tar xzf expat-2.1.0.tar.gz
cp -R expat-2.1.0/lib .
dng sdk编译
在xmp_sdk文件夹所在的目录下执行以下命令获取源码
git clone https://github.com/yanburman/dng_sdk.git
进入目录projects/linux/,执行make即可编译生成dng_validate,该工具用于校验dng文件,后面生成的文件可以通过该工具来检测。
bayer转dng编译
拷贝下面的代码并命名为bayer2dng.cpp,
#include <iostream>
#include "dng_color_space.h"
#include "dng_date_time.h"
#include "dng_exceptions.h"
#include "dng_file_stream.h"
#include "dng_globals.h"
#include "dng_host.h"
#include "dng_ifd.h"
#include "dng_image_writer.h"
#include "dng_info.h"
#include "dng_linearization_info.h"
#include "dng_mosaic_info.h"
#include "dng_negative.h"
#include "dng_preview.h"
#include "dng_render.h"
#include "dng_simple_image.h"
#include "dng_tag_codes.h"
#include "dng_tag_types.h"
#include "dng_tag_values.h"
#include "dng_xmp.h"
#include "dng_xmp_sdk.h"
#include "dng_camera_profile.h"
// --------------------------------------------------------------------------------
//
// MakeDNGSample
//
int main(int argc, char *argv[])
{
// Sample BAYER image at ISO100 and tEXP 1/10 on f/4.0 and focal length 10mm
std::string m_szInputFile = "pic6";
uint16 m_unISO = 100;
double m_dExposureTime = 0.1;
double m_dLensAperture = 4.0;
double m_dFocalLength = 10.0;
// SETTINGS: 12-Bit RGGB BAYER PATTERN
uint8 m_unColorPlanes = 3;
uint8 m_unBitDepth = 12;
uint16 m_unBayerType = 1; // RGGB
uint32 m_ulWidth = 1920;
uint32 m_ulHeight = 1225;
uint32 m_ulBlackLevel = 200;
// SETTINGS: Whitebalance D65, Orientation "normal"
dng_orientation m_oOrientation = dng_orientation::Normal();
dng_xy_coord m_oWhitebalanceDetectedXY = D65_xy_coord();
// SETTINGS: Names
std::string m_szMake = "My Company";
std::string m_szCameraModel = "My Camera";
std::string szProfileName = m_szCameraModel;
std::string szProfileCopyright = m_szMake;
// Calculate bit limit
uint16 m_unBitLimit = 0x01 << m_unBitDepth;
// Form output filenames
std::string szBaseFilename = "";
std::string m_szOutputFile = "";
std::string m_szRenderFile = "";
std::string m_szPathPrefixInput = "";
std::string m_szPathPrefixOutput = "";
std::string m_szPathPrefixProfiles = "";
size_t unIndex = m_szInputFile.find_last_of(".");
if (unIndex == std::string::npos)
{
szBaseFilename = m_szInputFile;
}
else
{
szBaseFilename = m_szInputFile.substr(0, unIndex);
}
m_szInputFile = m_szPathPrefixInput + m_szInputFile;
m_szOutputFile = m_szPathPrefixOutput + szBaseFilename + ".dng";
m_szRenderFile = m_szPathPrefixOutput + szBaseFilename + ".tiff";
// Create DNG
try
{
dng_host oDNGHost;
// -------------------------------------------------------------
// Print settings
// -------------------------------------------------------------
printf("\n");
printf("===============================================================================\n");
printf("Simple DNG converter\n");
printf("===============================================================================\n\n");
printf("\n");
printf("BAYER:\n");
printf("%s\n", m_szInputFile.c_str());
printf("\nConverting...\n");
// -------------------------------------------------------------
// BAYER input file settings
// -------------------------------------------------------------
dng_file_stream oBayerStream(m_szInputFile.c_str());
//oBayerStream.SetLittleEndian(false);
uint32 ullBayerSize = (uint32)oBayerStream.Length();
AutoPtr<dng_memory_block> oBayerData(oDNGHost.Allocate(ullBayerSize));
oBayerStream.SetReadPosition(0);
oBayerStream.Get(oBayerData->Buffer(), ullBayerSize);
// -------------------------------------------------------------
// DNG Host Settings
// -------------------------------------------------------------
// Set DNG version
// Remarks: Tag [DNGVersion] / [50706]
oDNGHost.SetSaveDNGVersion(dngVersion_SaveDefault);
// Set DNG type to RAW DNG
// Remarks: Store Bayer CFA data and not already processed data
oDNGHost.SetSaveLinearDNG(false);
// -------------------------------------------------------------
// DNG Image Settings
// -------------------------------------------------------------
dng_rect vImageBounds(m_ulHeight, m_ulWidth);
AutoPtr<dng_image> oImage(oDNGHost.Make_dng_image(vImageBounds, m_unColorPlanes, ttShort));
dng_pixel_buffer oBuffer;
oBuffer.fArea = vImageBounds;
oBuffer.fPlane = 0;
oBuffer.fPlanes = 1;
oBuffer.fRowStep = oBuffer.fPlanes * m_ulWidth;
oBuffer.fColStep = oBuffer.fPlanes;
oBuffer.fPlaneStep = 1;
oBuffer.fPixelType = ttShort;
oBuffer.fPixelSize = TagTypeSize(ttShort);
oBuffer.fData = oBayerData->Buffer();
oImage->Put(oBuffer);
// -------------------------------------------------------------
// DNG Negative Settings
// -------------------------------------------------------------
AutoPtr<dng_negative> oNegative(oDNGHost.Make_dng_negative());
// Set camera model
// Remarks: Tag [UniqueCameraModel] / [50708]
oNegative->SetModelName(m_szCameraModel.c_str());
// Set localized camera model
// Remarks: Tag [UniqueCameraModel] / [50709]
oNegative->SetLocalName(m_szCameraModel.c_str());
// Set bayer pattern information
// Remarks: Tag [CFAPlaneColor] / [50710] and [CFALayout] / [50711]
oNegative->SetColorKeys(colorKeyRed, colorKeyGreen, colorKeyBlue);
// Set bayer pattern information
// Remarks: Tag [CFAPlaneColor] / [50710] and [CFALayout] / [50711]
oNegative->SetBayerMosaic(m_unBayerType);
// Set bayer pattern information
// Remarks: Tag [CFAPlaneColor] / [50710] and [CFALayout] / [50711]
oNegative->SetColorChannels(m_unColorPlanes);
// Set linearization table
// Remarks: Tag [LinearizationTable] / [50712]
AutoPtr<dng_memory_block> oCurve(oDNGHost.Allocate(sizeof(uint16)*m_unBitLimit));
for ( int32 i=0; i<m_unBitLimit; i++ )
{
uint16 *pulItem = oCurve->Buffer_uint16() + i;
*pulItem = (uint16)(i);
}
oNegative->SetLinearization(oCurve);
// Set black level to auto black level of sensor
// Remarks: Tag [BlackLevel] / [50714]
oNegative->SetBlackLevel(m_ulBlackLevel);
// Set white level
// Remarks: Tag [WhiteLevel] / [50717]
oNegative->SetWhiteLevel(m_unBitLimit-1);
// Set scale to square pixel
// Remarks: Tag [DefaultScale] / [50718]
oNegative->SetDefaultScale(dng_urational(1,1), dng_urational(1,1));
// Set scale to square pixel
// Remarks: Tag [BestQualityScale] / [50780]
oNegative->SetBestQualityScale(dng_urational(1,1));
// Set pixel area
// Remarks: Tag [DefaultCropOrigin] / [50719]
oNegative->SetDefaultCropOrigin(0, 0);
// Set pixel area
// Remarks: Tag [DefaultCropSize] / [50720]
oNegative->SetDefaultCropSize(m_ulWidth, m_ulHeight);
// Set base orientation
// Remarks: See Restriction / Extension tags chapter
oNegative->SetBaseOrientation(m_oOrientation);
// Set camera white XY coordinates
// Remarks: Tag [AsShotWhiteXY] / [50729]
oNegative->SetCameraWhiteXY(m_oWhitebalanceDetectedXY);
// Set baseline exposure
// Remarks: Tag [BaselineExposure] / [50730]
oNegative->SetBaselineExposure(0);
// Set if noise reduction is already applied on RAW data
// Remarks: Tag [NoiseReductionApplied] / [50935]
oNegative->SetNoiseReductionApplied(dng_urational(0,1));
// Set baseline sharpness
// Remarks: Tag [BaselineSharpness] / [50732]
oNegative->SetBaselineSharpness(1);
// -------------------------------------------------------------
// DNG EXIF Settings
// -------------------------------------------------------------
dng_exif *poExif = oNegative->GetExif();
// Set Camera Make
// Remarks: Tag [Make] / [EXIF]
poExif->fMake.Set_ASCII(m_szMake.c_str());
// Set Camera Model
// Remarks: Tag [Model] / [EXIF]
poExif->fModel.Set_ASCII(m_szCameraModel.c_str());
// Set ISO speed
// Remarks: Tag [ISOSpeed] / [EXIF]
poExif->fISOSpeedRatings[0] = m_unISO;
poExif->fISOSpeedRatings[1] = 0;
poExif->fISOSpeedRatings[2] = 0;
// Set WB mode
// Remarks: Tag [WhiteBalance] / [EXIF]
poExif->fWhiteBalance = 0;
// Set metering mode
// Remarks: Tag [MeteringMode] / [EXIF]
poExif->fMeteringMode = 2;
// Set metering mode
// Remarks: Tag [ExposureBiasValue] / [EXIF]
poExif->fExposureBiasValue = dng_srational(0, 0);
// Set aperture value
// Remarks: Tag [ApertureValue] / [EXIF]
poExif->SetFNumber(m_dLensAperture);
// Set exposure time
// Remarks: Tag [ExposureTime] / [EXIF]
poExif->SetExposureTime(m_dExposureTime);
// Set focal length
// Remarks: Tag [FocalLength] / [EXIF]
poExif->fFocalLength.Set_real64(m_dFocalLength, 1000);
// Set lens info
// Remarks: Tag [LensInfo] / [EXIF]
poExif->fLensInfo[0].Set_real64(m_dFocalLength, 10);
poExif->fLensInfo[1].Set_real64(m_dFocalLength, 10);
poExif->fLensInfo[2].Set_real64(m_dLensAperture, 10);
poExif->fLensInfo[3].Set_real64(m_dLensAperture, 10);
// -------------------------------------------------------------
// DNG Profile Settings: Simple color calibration
// -------------------------------------------------------------
// Camera space RGB to XYZ matrix with D65 illumination
// Remarks: Derived from MATLAB using least square linear regression with
// MacBeth ColorChecker classic
dng_matrix_3by3 oCameraRGB_to_XYZ_D65 = dng_matrix_3by3( 2.3150, 0.0711, 0.1455,
0.9861, 0.7815,-0.2192,
0.2082,-0.3349, 1.6432 );
uint32 ulCalibrationIlluminant1 = lsD65;
// Camera space RGB to XYZ matrix with StdA illumination
// Remarks: Derived from MATLAB using least square linear regression with
// MacBeth ColorChecker classic
dng_matrix_3by3 oCameraRGB_to_XYZ_A = dng_matrix_3by3( 1.6335, 0.4718,-0.0656,
0.5227, 1.0298,-0.3416,
0.0475,-0.2020 ,1.2522 );
uint32 ulCalibrationIlluminant2 = lsStandardLightA;
AutoPtr<dng_camera_profile> oProfile(new dng_camera_profile);
// Set first illuminant color calibration if available
if ( ulCalibrationIlluminant1!=0 )
{
// Set calibration illuminant 1
// Remarks: Tag [CalibrationIlluminant1] / [50778]
oProfile->SetCalibrationIlluminant1(ulCalibrationIlluminant1);
// Set color matrix 1
// Remarks: Tag [ColorMatrix1] / [50721]
oProfile->SetColorMatrix1(Invert(oCameraRGB_to_XYZ_D65));
}
// Set second illuminant color calibration if available
if ( ulCalibrationIlluminant2!=0 )
{
// Set calibration illuminant 2
// Remarks: Tag [CalibrationIlluminant2] / [50779]
oProfile->SetCalibrationIlluminant2(ulCalibrationIlluminant2);
// Set color matrix 1
// Remarks: Tag [ColorMatrix2] / [50722]
oProfile->SetColorMatrix2(Invert(oCameraRGB_to_XYZ_A));
}
// Set name of profile
// Remarks: Tag [ProfileName] / [50936]
oProfile->SetName(szProfileName.c_str());
// Set copyright of profile
// Remarks: Tag [ProfileCopyright] / [50942]
oProfile->SetCopyright(szProfileCopyright.c_str());
// Force flag read from DNG to make sure this profile will be embedded
oProfile->SetWasReadFromDNG(true);
// Set policy for profile
// Remarks: Tag [ProfileEmbedPolicy] / [50941]
oProfile->SetEmbedPolicy(pepAllowCopying);
// Add camera profile to negative
oNegative->AddProfile(oProfile);
// -------------------------------------------------------------
// Write DNG file
// -------------------------------------------------------------
// Assign Raw image data.
oNegative->SetStage1Image(oImage);
// Compute linearized and range mapped image
oNegative->BuildStage2Image(oDNGHost);
// Compute demosaiced image (used by preview and thumbnail)
oNegative->BuildStage3Image(oDNGHost);
// Update XMP / EXIF
oNegative->SynchronizeMetadata();
// Update IPTC
oNegative->RebuildIPTC(true);
// Create stream writer for output file
dng_file_stream oDNGStream(m_szOutputFile.c_str(), true);
//oDNGStream.SetSwapBytes(1);
// Write DNG file to disk
AutoPtr<dng_image_writer> oWriter(new dng_image_writer());
oWriter->WriteDNG(oDNGHost, oDNGStream, *oNegative.Get(), NULL, ccUncompressed);
#if 0
// -------------------------------------------------------------
// Write TIFF file
// -------------------------------------------------------------
// Create stream writer for output file
dng_file_stream oTIFFStream(m_szRenderFile.c_str(), true);
// Create render object
dng_render oRender (oDNGHost, *oNegative);
// Set exposure compensation
oRender.SetExposure(0.0);
// Create final image
AutoPtr<dng_image> oFinalImage;
// Render image
oFinalImage.Reset (oRender.Render ());
oFinalImage->Rotate(oNegative->Orientation());
// Write TIFF file to disk
oWriter->WriteTIFF(oDNGHost, oTIFFStream, *oFinalImage.Get(), piRGB, ccUncompressed, oNegative.Get(), &oRender.FinalSpace ());
#endif
}
catch (const dng_exception &except)
{
return except.ErrorCode ();
}
catch (...)
{
return dng_error_unknown;
}
printf ("Conversion complete\n");
return 0;
}
保存,然后执行下面命令编译
g++ bayer2dng.cpp -I../../../xmp_sdk/public/include -I../../source libdng_sdk.a -ljpeg -lz -lpthread -ldl ../../../xmp_sdk/public/libraries/i80386linux_x64/release//staticXMPFiles.ar ../../../xmp_sdk/public/libraries/i80386linux_x64/release//staticXMPCore.ar -ldl -o bayer2dng
即可生成bayer转rgb的工具bayer2dng了。
总结
上面的bayer2dng.cpp仅仅是将1920 1225分辨率 12bit RGGB格式的raw转为dng格式的文件,里面很多参数都需要根据实际情况调整,比如曝光时间、iso、光圈大小、焦距、线性表等等,所以继续研究dng格式是避免不了的_
引用
- Supported camera models
- MRW File Format
- How to get Raw data from CR2 format.
- Understanding What is stored in a Canon RAW .CR2 file, How and Why
- Digital negative
- MRW Camera Raw Signature Format: Specification & MRW Recovery Example
- Canon's CR2 Raw File Format Specification
- The Canon RAW (CRW) File Format
- Digital Cameras - Canon EOS-1D X Test Images
- TIFF图像文件格式详解
- TIFF
- TIFF图像文件格式详解
- TIFF Tag Reference
- Intel格式与Motorola格式的区别
- SubIFD1 tags not updating
- How to import a real raw CFA image into a DNG ?
- ChromaSoft: dcpTool and the DNG SDK on Linux
- DNG tools in LINUX | shalamander
- dcraw-float command line options
完!
2016年12月
bayer转dng实现过程记录的更多相关文章
- 升级Windows 10 正式版过程记录与经验
升级Windows 10 正式版过程记录与经验 [多图预警]共50张,约4.6MB 系统概要: 预装Windows 8.1中文版 64位 C盘Users 文件夹已经挪动到D盘,并在原处建立了符号链接. ...
- 双系统Ubuntu分区扩容过程记录
本人电脑上安装了Win10 + Ubuntu 12.04双系统.前段时间因为在Ubuntu上做项目要安装一个比较大的软件,导致Ubuntu根分区的空间不够了.于是,从硬盘又分出来一部分空间,分给Ubu ...
- CentOS 5.5 下安装Countly Web Server过程记录
CentOS 5.5 下安装Countly Web Server过程记录 1. 系统更新与中文语言包安装 2. 基本环境配置: 2.1. NodeJS安装 依赖项安装 yum -y install g ...
- linux-i386(ubuntu)下编译安装gsoap_2.8.17过程记录
过程记录 : 1.下载gsoap_2.8.17.zip 并 解压 : $unzip gsoap_2.8.17.zip 2.进入解压后的目录gsoap-2.8 3.自动配置编译环境: $ ...
- 【转】android 最新 NDK r8 在window下开发环境搭建 安装配置与使用 详细图文讲解,完整实际配置过程记录(原创)
原文网址:http://www.cnblogs.com/zdz8207/archive/2012/11/27/android-ndk-install.html android 最新 NDK r8 在w ...
- 升级到 ExtJS 5的过程记录
升级到 ExtJS 5的过程记录 最近为公司的一个项目创建了一个 ExtJS 5 的分支,顺便记录一下升级到 ExtJS 5 所遇到的问题以及填掉的坑.由于 Sencha Cmd 的 sencha ...
- Ubuntu14.04 Tomcat 安装过程记录
Ubuntu14.04 Tomcat 安装过程记录 检查java的版本 zhousp@ubuntu:~$ sudo java -version [sudo] password for zhousp: ...
- mercurial(Hg) Server 搭建 过程记录
mercurial(Hg) Server 搭建 过程记录 1. 环境说明 只是测试搭建,环境为本机开发环境:win 8.1 + IIS8.5 软件准备: 2. 软件安装 先安装Python2.7, ...
- xp硬盘安装Fedora14 过程记录及心得体会(fedora14 live版本680M 和fedora14 DVD版本3.2G的选择)
这次电脑奔溃了,奇怪的是直接ghost覆盖c盘竟然不中.之前电脑上硬盘安装的fedora14操作系统,也是双系统.不知道是不是这个问题,记得同学说过,在硬盘装fedora之后,要手动修改c盘隐藏的那个 ...
随机推荐
- selenide 自动化UI测试中Configuration全局配置项目
selenide 在测试过程中需要设置许多的默认值,方便在测试过程中进行和很好的使用.下面我们在selenide中的api引用过来看看! static Configuration.AssertionM ...
- 第三十篇 面向对象的三大特性之继承 supre()
继承 一 .什么是继承? 类的继承跟现实生活中的父.子.孙子.重孙子的继承关系一样,父类又称基类. Python中类的继承分为:单继承 和 多继承. # 定义父类 class ParentClass ...
- kaldi解码及特征提取详解
目录 1. 注意事项 2. 流程图: 3. 具体流程指令: 1. 注意事项 首先要训练好模型,用到3个文件,分别是: final.mdl(训练模型得到的模型文件) final.mat(用来特征转换) ...
- oop &&GP 模板 ---> 特化和偏特化
OOP面向对象编程 GP泛型编程(generic programming) 两者的主要区别就是OOP将数据和对数据的操作放在一起, GP就是将数据和操作独立开来 GP: 数据就是container ...
- iOS-调用百度地图,苹果自带地图,高德地图,谷歌地图导航方法
- (void)actionSheet : (ServiceNetworkModel *)model{ __block NSString *urlScheme = @"demoURI://& ...
- Java中IO——NIO
一.引入 当引入一些新功能的时候,那说明之前的设计可能还需要完善. 1.阻塞式 在传统的IO输入输出中,如果我们从流中去读数据,而数据源中没有数据时,程序就会阻塞该线程.阻塞式线程的一种基本状态,可以 ...
- vue2.0中vue-router使用总结
#在vue-cli所创建的项目中使用 进入到项目的目录后使用 npm install vue-router --save 安装vue-router,同时保存在webpack.Json配置文件中,然 ...
- Eclipse打不开,闪退
自己编写了个程序,运行巨慢..无语,输出太多,后来冒出一个错误,不知什么原因啊,再后来Eclipse就打不开了,到workbench闪退... 百度后解决方案: 进入目录:workspace/.met ...
- 【题解】洛谷P2418 yyy loves OI IV
感觉很是妙啊……这题数次误入歧途...最开始想的二维dp,单调队列优化:无果,卒.于是没忍住看了下标签:暴力枚举?搜索?于是开始想记忆化搜索.以为会有什么很强的剪枝之类的:30分,卒.最后终于回到正道 ...
- [poj 3252]数位dp前导0的处理
通过这个题对于数位dp中前导0的处理有了新的认识. 题目链接:http://poj.org/problem?id=3252 //http://poj.org/problem?id=3252 #incl ...