fengbingchun / opencv_test

fbc_cv is an open source image process library.

OpenCV_Test

The main role of the project:

• OpenCV's usage OpenCV GitHub
• fbc_cv library: an open source image process library
• libyuv's usage libyuv GitHub
• VLFeat's usage vlfeat.org
• Vigra's usage vigra GitHub
• CImg's usage cimg.eu
• FFmpeg'usage ffmpeg.org
• LIVE555'usage LIVE555.COM
• libusb'usage libusb GitHub
• libuvc'usage libuvc GitHub
• libexif'usage libexif GitHub

The version of each open source library see: version.txt

The project support platform:

• windows10 64 bits: It can be directly build with VS2022 in windows10 64bits.
• Linux:
  • OpenCV_Test support cmake build(file position: prj/linux_cmake_OpenCV_Test)
  • FFmpeg_Test support cmake build(test code include: FFmpeg, LIVE555; file position: prj/linux_cmake_FFmpeg_Test)

OpenCV's version: 3.1

• close support for OpenCL/CUDA/SIMD/TBB/OpenMP when build with CMake
• modify modules/core/include/opencv2/core/cvdef.h,close SIMD support: adjust line 167 to: #if 0
• in order to keep the linux and windows results consistent, modify modules/core/include/opencv2/core/fast_math.hpp: make cvFound/cvFloor/cvCeil/cvRound to execute the last branch
• insure that all algorithms are implemented with c++

fbc_cv

• it is an open source image process library
• most of the algorithms come from OpenCV3.1
• it has a template class Mat_ replace of OpenCV's Mat class: src/fbc_cv/include/core/mat.hpp
• interface names are consistent with OpenCV3.1
• each algorithm's result is same with OpenCV3.1
• the codes are written in C++ without dependence on any 3rd-party libraries

fbc_cv have been implemented include:

• resize
• cvtColor
• merge/split
• remap
• warpAffine
• rotate
• warpPerspective
• dilate
• erode
• morphologyEx
• threshold
• transpose
• flip
• dft/idft
• Windows DShow, support codec type: mjpeg
  • opencv: 2.4.13.6, only get decoded data
  • ffmpeg: 4.2, only get encoded data

Python code(support windows and linux platform):

• environment configuration
  • install anaconda
  • install imagemagick
  • execute the following command:

     conda create -n OpenCV_Test python=3.9
     conda activate OpenCV_Test
     git clone https://github.com/fengbingchun/OpenCV_Test
     cd OpenCV_Test/demo/Python
     pip install -r requirements.txt
    

  • modify Anaconda3/envs/OpenCV_Test/Lib/site-packages/moviepy/config_defaults.py

     IMAGEMAGICK_BINARY = os.getenv('IMAGEMAGICK_BINARY', 'auto-detect') # original statement
     IMAGEMAGICK_BINARY = os.getenv('IMAGEMAGICK_BINARY', 'the magick executable for the actual path, for example: C:\Program Files\ImageMagick-7.1.1-Q16-HDRI\magick.exe') # modified statement
    

• features
  • camera calibration
  • add text or image to the video
  • image generic operations

Screenshot:

Blog:

• opencv/fbc_cv
• ffmpeg/live555

fbc_cv library licence: uses the same licence as OpenCV3.1

opencv_test's Issues

resize算法求解

大佬请问一下你实现的 U8 的 resize 算法和 opencv 能做到 bit对齐吗？

dft.hpp代码src = dst过不了编译

很棒！但是昨天我在vs2017编译时保了以下错误：
C2679 二进制“=”: 没有找到接受“fbc::Mat_<float,2>”类型的右操作数的运算符(或没有可接受的转换)
似乎是vs的标准不允许chs不同的模板类之间的直接赋值，目前为止没有在2013上测试过，所以这可能是140+的限制？有没有什么办法避开这个问题呢？

为什么插值算法的nearest neighbour这么复杂 对比之前的

不好意思，打扰了
https://github.com/fengbingchun/OpenCV_Test/blob/master/src/fbc_cv/include/resize.hpp
line 844-932 里面有不同case
简单的方法：https://blog.csdn.net/fengbingchun/article/details/17335477 这里的resizeNN很简单
这两种方法 有什么不同？

在Ubuntu上编译的无法读取图像

在Ubuntu上编译的时候遇到以下几个问题：

1. src/fbc_cv/include/core 文件夹下的头文件有含有"core/XXX"的，同一个文件夹下不需要加"core/...."吧，不得不挨个删掉"core/"，不知是否有更好的解决方法.

2. 无法读取图像的错误

原因是test_opencv_funset.cpp的18行，在linux平台下读取了

matSrc = cv::imread("lena.jpg");

should:

matSrc = cv::imread("test_images/lena.png");