Home » Programming » Image Processing
Category:

Image Processing

ContentImage ProcessingOpenCVProgramming

Saving OpenCV matrices to disc and loading them

by Jack Simpson
written by Jack Simpson

Recently I needed to save some matrices I had generated with OpenCV and C++. This snippet of code shows you how you can do this this and then open the files to retrieve the matrices.  You can save the files as either the .yml or .xml format. In the example below, “trainData” and “trainLabels” are the two matrices I want to save:

cv::FileStorage file("tag_data.yml", cv::FileStorage::WRITE);

file << "data" << trainData;
file << "classes" << trainLabels;
file.release();

// later (in another program when you want to get the stored matrices)

cv::Mat trainDataCopy, trainLabelsCopy;

cv::FileStorage file("tag_data.yml", cv::FileStorage::READ);

file["data"] >> trainDataCopy;
file["classes"] >> trainLabelsCopy;

file.release();

Done. You should now be able to use both matrices.

0 comment
0 FacebookTwitterPinterestEmail
ContentImage ProcessingResearchTips & Tutorials

Be careful with FFmpeg metadata

by Jack Simpson
written by Jack Simpson

Recently I used the program FFmpeg to automatically extract metadata about the time and date some videos I’ve been using for my research were created. FFmpeg is a really useful tool for manipulating videos and images at the command-line. You can do things like change the format, quality or length of your video files with relative ease. The image processing library OpenCV even uses FFmpeg under the hood for opening videos you want to process. You can also use FFmpeg to extract metadata about your video files, and that is what I was using it for. However, I soon discovered that the metadata I retrieved with the program wasn’t completely accurate. Video file creation times were off by a couple of hours. Normally that doesn’t matter much, however if you’re breaking your data down hour-by-hour over a 2 week period, having these kinds of discrepancies really throws your analysis out. It was a great reminder to always thoroughly check the output of your analysis pipelines and to never assume that a tool will work perfectly all the time.

0 comment
0 FacebookTwitterPinterestEmail
ContentData ScienceImage ProcessingMachine LearningOpenCV

Image processing with scikit-image

by Jack Simpson
written by Jack Simpson

I’ve been using OpenCV over the past couple of years for all my image processing work. It’s a really extensive and useful library, and just recently, OpenCV 3.0 was released, which included bindings for Python 3! Finally one of my last reasons to continue using Python 2 has disappeared. However, there is another image processing library that I just found out about – scikit-image. Yes, that’s the same scikit as the one from scikit-learn, the awesome machine learning library for Python that I’d linked to previously in my machine learning resources post. While I plan to continue using OpenCV, I had a read through some of the scikit-image tutorials with interest, and if you’re looking for a light-weight library for image processing that plays well with SciPy, I’d definitely check it out.

If you’re interested in learning more about image processing with OpenCV, I’m preparing a presentation for the Canberra Python Users Group using the iPython Notebook which provides an introduction to using this library. I’ll upload the notebook to GitHub and write a post linking to it in a few weeks.

0 comment
0 FacebookTwitterPinterestEmail
ContentImage ProcessingOpenCV

Review of OpenCV Essentials

by Jack Simpson
written by Jack Simpson

Recently PACT sent me a free copy of a book on OpenCV because of my background writing about the topic. First things first, OpenCV Essentials is definitely not a book for an image processing/computer vision novice. If this sounds like you, then I’d recommend you refer to my post on resources for getting started with OpenCV here, and then come back to this book in a few months. However, if you’ve been using OpenCV and programming in C++ for at least a few months and want a resource which you can quickly refer to when you’re delving into some of the deeper functionality of OpenCV, then I can thoroughly recommend this book to you.

Continue Reading
0 comment
0 FacebookTwitterPinterestEmail
ContentImage ProcessingOpenCVProgramming

Best resources for learning OpenCV (Python and C++)

by Jack Simpson
written by Jack Simpson

Here’s a list of some of the websites, books and videos (for both Python and C++), that I have found very useful while learning OpenCV.

Python

  1. Using OpenCV with Python and ROS (Video) – Great introduction to computer vision and OpenCV.
  2. PyImageSearch – Website has a lot of great tutorials on many different applications of OpenCV.
  3. OpenCV-Python Tutorials (Website) – Official Python documentation and tutorials.
  4. OpenCV-Python Tutorials (Website) – Comprehensive set of tutorials.
  5. OpenCV Python (Website) – The original Python tutorials I started with, most of these have been ported to the website above.
  6. OpenCV 3 With Python (Website) – Great list of Python tutorials.
  7. Tentative NumPy Tutorial (Website) – As OpenCV reads images into NumPy arrays, its useful to have an understanding of this library.
  8. Python – Getting started with OpenCV (Website) – Introduction to using Python with OpenCV.
  9. Programming Computer Vision with Python (Book) – How to use Python to build computer vision programs.
  10. OpenCV Computer Vision with Python (Book) – Short book on getting started with OpenCV and Python.
  11. Building an image processing pipeline with Python (Video) – Talk about how Python and OpenCV are used in industry.
Continue Reading
0 comment
0 FacebookTwitterPinterestEmail
BloggingContentImage ProcessingOpenCVTips & Tutorials

Using C++ or Python for OpenCV Programs

by Jack Simpson
written by Jack Simpson

When I first started learning OpenCV, I was working exclusively with Python. While I am still a huge fan of the language, today all of my OpenCV programs are written in C++. Why?

  1. Some of the deeper functionality of OpenCV has not been completely ported to Python (although hopefully the release of OpenCV 3.0 will fix most of these issues).
  2. Most in-depth textbooks on image processing and computer vision that cover OpenCV use C++ as their primary language. It was therefore easier to learn from these resources by adopting the language.
  3. A lot of the computer vision techniques I use (SIFT, machine learning, etc), are better documented in C++.
  4. Passing images back and forth between NumPy arrays has overhead that C++ doesn’t have to worry about.
  5. As has been suggested to me by Carl Bell, Python struggles to perform well with overloaded functions.
Continue Reading
0 comment
0 FacebookTwitterPinterestEmail
ContentOpenCVProgrammingTips & Tutorials

Sublime Text OpenCV C++ Build System

by Jack Simpson
written by Jack Simpson

If you want to compile and run a C++ program using OpenCV in Sublime Text, then copy and paste the code below into a build system file. If you’re not interested in explicitly using C++11 you can delete the “-std=c++0x” section.


{
"shell_cmd": "g++ -std=c++0x ${file} -o ${file_base_name} `pkg-config --cflags --libs opencv` && ./${file_base_name}"
}
0 comment
0 FacebookTwitterPinterestEmail
ContentImage ProcessingOpenCVResearch

Tracking bees using local features

by Jack Simpson
written by Jack Simpson

The project I am working on at present is focused on building an automated system for tracking the movement of honeybees in an observation hive, filmed using a grayscale camera and infrared light. My recent attempt at this has involved extracting features (or keypoints) from the frames and then describing them using the SIFT (Scale Invariant Feature Transform) algorithm, implemented in the OpenCV imaging library. There are a number of other (potentially faster) feature detection algorithms available (SURF, ORB, etc), however as SIFT has been traditionally regarded as the most robust, I thought I would start with it. SIFT works first by extracting what it deems to be robust keypoints/features. For each of these, it will then compute a unique feature descriptor based on the 16×16 pixel area of the feature. By extracting and describing these features, you can try to compare images as shown below.

Continue Reading
0 comment
0 FacebookTwitterPinterestEmail