Tag Archives: librawprocessor

CLAHE shading correction

Applying CLAHE to medical images may going to some shaded on each edge side of object. So I have to correct this problem with fast processing in real time.

To make up a prototype correction, I have used same source RAW image, here is a source image in down scaled from 16bit gray scale RAW.

Applied CLAHE with 16×16 in clip limit of 100.0f. And I could recognized shaded areas at each edge of object. Shaded area marked as red boxes in below.

To correct shaded area, generate gaussian blur mask with big radials, but it really heavy to processing in fast time as realtime. So I made it fastest way with my ‘Resize engine’ that made with OpenMP and AVX instructions. To generate fastest blurred mask, down scale with bi-linear filter to 2.5% size. Then doing up scale again with B-Spline filter to original size. And Invert it.

Then calculate to do shade correction with generated shade mask, ins fastest math functions with original image.

To complete image processing, need to fill background areas.

It simply corrected but little bit lacks on details of bone level. But definitely better than hard shaded levels after window leveling. Expect for next will find more improved processing algorithm, and it will be a function of librawprocessor.

tinydicom + rawprocessor lesson #2

Now it’s time to next step for using my open sources,  And here I like to introduce how loads DCM and what it contains inside tags.

Just read my example code to read DICOM tags from DCM file, it should help what inside.

It may compiles with libtinydicom.a linkage as well, just compile your code with -ltinydicom -L{libtinydicom.a directory from your path}. Or, just follow my way – Make an empty directory in your working directory –  I like to recommend to make a directory as like ‘projects’. Then make a new again for unique name to check for next time, or continue to more works : recommend ‘dcmtest1’ or ‘dcmtest2’. So you may now placed to {Your home directory}/projects/dcmtest1 when you had followed my recommendation. Now make a empty cpp file with any editor like vim, then copy & paste my example, then save it up and quit editor. Then copy prebuilt “libtinydicom.a” and “libdcm.h” into your current directory – if you didn’t prebuilt? just build it first. and type like this.

g++ main.cpp -ltinydicom -o test1

You may check test1 or test1.exe (on Windows).
Now you need download sample DCM file from here:

SMPTE & SIMFIT

If you have wget, you can do :

wget http://tomintechsupport.com/download/SIMFIT.dcm

Now you have SIMFIT.dcm and compiled binary with same directory. Just type your binary name. And check result. 87 items may displayed with sizes.

 

Then, try to next step, save raw pixel image to a new file. See my example source.

Now you need prebuilt librawprocessor.a and rawprocessor.h file in your directory with example source code. It proceeds read SIMFIT.DCM and export pixel image information, then writes to a file.

Important points is ‘Checking Window center and width’ from DICOM tag ID 0028:1050 and 0028:1051. These tags contains information of window center and width, and it must applied to exported image. So I used Get16bitThresholdedImage() method to make windowed image.

If you need more detailed? just let left a guestbook here.

Next lesson may using libPNG, and make raw image to 8bit grey scaled image to see in common image viewing programs.

tinydicom + rawprocessor lesson #1

Here I planned to write some lessons to understand what is libtinydicom (Tiny DICOM library) and librawprocessor (RAW Processing library) for most of modern compilers except M$VC – unfortunately I am sorry about supporting M$VC, but I don’t like care about distinct, banishing standard compiler.

Ok, let understand what is each library doing what it does and what it existed for.

libtinydicom

This library was born to read and write DICOM tags in a file by myself. At a time I made this, there’s some commercial libraries and open source too – A.K.A DCMTK. I was just wanted to write simply some DICOM tags and also reads Pixel datas from any DCM files – so I have started to read NEMA’s DICOM standards. And it was about 2011.

It passed years and continuously made it works well with many different DCM files, and now it availed to as an open source on GitHub.com.

Anyone can clone or download source code and use it for freely with small license MIT. Just Ok for announcing what you used it your project or product.

Supported compilers are most of modern C++ compiler supporting Makefile. It made with MinGW-W64 and Code::Blocks IDE, but now supporting Mac OS-X and Linux too because it made only with C++ STL.

How to use it ?

Simply you can download source code as ZIP compression from here. And you may need already know how I can extract ZIP file to somewhere, and plus more you must know what I have to use compiler with this library – If don’t know? Just, please, don’t try to use it, It may difficult to understand next jobs.

Now choose one of Makefiles that have extensions – gcc and llvm.  Just copy one Makefile.{your compiler to use} to Makefile as like cp Makefile.gcc Makefile and then, just type to make.   If it succeed to build all sources, you can see a lib directroy, and there’s two files : libtinydicom.a and libdcm.h

And now you may understand how it be used, and this lesson may continued to next.

 

librawprocessor

librawprocessor born to doing something for most of medical RAW images. It supports read and write, rotating, flip, convert to 8 bit image and more features.

You can download or clone source codes here.

“librawprocessor” supports OpenMP for processing image quickly if your compiler supports this feature. Most of processing features are optized for OpenMP and AVX instruction.

Features:

  • Load from File, Memory.
  • Save as a new RAW image.
  • Flip vertical, horizontal
  • Fast Rotate 90, 180, 270 degrees
  • Free Rotate in 0 to 359.99 degrees
  • Change width and height in size of height, to recognize what resolution is right.
    ( Most medical RAW images has no information about sizes )
  • Invert
  • Rescale with many filters : Bilinear, Bicubic, B-Spline, Lanzcos
  • Kernel matrix filtering : sharpen, blur or customized
  • Brightness, Contrast, Gamma adjustment
  • Tone mapping with Drago and Reinhard
  • CLAHE
  • Generate low frequency image and anisotropic filtered image from source
  • Advanced edge enhace
  • Automatic weight calculation ( window width )
  • Get down scaled pixel ranged data from source ( thresholding to any bits, 8 or 16bits )

How to use it ?

Simply you can download or clone from here. You may need prepared to understand what is your compiler and how to build with Makefile. ( M$VC not an option, sorry about this but no plan to future, too ) Tested almost of GCC environments : Windows, Mac OS-X, Linux.

After you got source code on your system, copy your right Makefile.{compiler type you have} to Makefile as like this : cp Makefile.gcc Makefile , then proceed simply make. And you can decide to use OpenMP option with appends a word openmp to make.

Required OpenMP is at least 3.0 or above. lower version may reason of compile failure.

And now you may understand how it be used, and this lesson may continued to next, too.

Applying HDR in medical image.

Prologue

High dynamic tone mapping is a kind of graphical thesis to indemnify exposure of whole a image specially such as 256 leveled Red, Green, Blue (+Alpha transparency) formats. But high dynamic calculates all pixel levels as an floating point number with luminance ( in case of RGB, it convert each color channels to a luminance level ) to enhance for more dynamic ranged.

Programming

Examined all algorithms and refered to Free Image 3 library for how to make it as C++ code. Proceeded to stand-alone codes and finally embedded to my open source project, librawprocessor at my github repository.

Testing results

As my experimental study, High dynamic tone mapping was enhances low exposed/qulity medical images to fully ranged pixel levels in same min/max range.

Left image is original digital medical chest PA TFT detector image and it could be ranged about 0 to 2700.

Right image is processed Reinhard alogortihm (with parameters : contrast 1.0 and adaptation 0.5) to expand dynamic range.

Pixel levels spreads down to about 6500 without high peak as above image. Every pixel levels be flatten. It will be expected to better for adjust post image processing with less loses.

Also it much better to examine anatomy with no special signal processing. Just doing thresholding window leveling makes good result to check organs.

Used program: