mmmn143 / fast-mser

Fast MSER and other comparison MSER algorithms

Fast MSER

Future Work

• Experiments w.r.t. different threads (1, 2, 4, 8, 16, 32 threads). We have a computer with 32-cores. The OS is Linux. Thus, we need to support our algorithms on Linux;

• Improve the speed-up in MSER recognition, and further reduce the running memory of V1 and V2;

• Compare Fast MSER to several deep learning based scene text methods in execution time aspect.

News (2020/05/02)

• We support our algorithms on MacOS (however, OpenMP is not supported well in MacOS);

• Fast MSER V1 supports 1, 2, 4, 8, 16 and 32 threads. Fast MSER V2 supports 1, 4 and 32 threads.

Algorithms and Codes

All codes are in the directory of basicimg.

OpenCV MSER (CV-MSER): img_opencv3_mser (original implementation https://github.com/opencv/opencv)

CV-MSER+: img_linked_linear_mser

VLFeat MSER (VF-MSER): img_vlfeat_mser (original implementation http://www.vlfeat.org/)

Idiap MSER (ID-MSER): img_idiap_mser (original implementation https://github.com/idiap/mser)

Fast-MSER V1: img_fast_mser_v1 (supporting 1-threads, 2-threads, 4-threads, 8-threads, 16-threads and 32-threads)

Fast MSER V2: img_fast_mser_v2 (supporting 1-threads, 4-threads and 32-threads)

Note that for channel parallel algorithms (CPCV-MSER, CPCV-MSER+, CPVF-MSER, CPID-MSER), we can use the following codes:

img_mser_base* mser = new img_idiap_mser(); // or other mser algorithms

#pragma omp parallel for num_threads(4) // 4 denotes the number of threads

for (i32 i = 0; i < 4; ++i) {

    img_multi_msers mser_res;

    mser->extract(mser_res, srcs[i]);

}

Examples of Runing an MSER algorithm

You can find the examples of how to run a MSER algorithm in the code: basicimg/test/img_mser_test.cpp

Performance w.r.t. Different deltas

Compared to CV-MSER+, the speed-ups of Fast MSER V1 w.r.t. different deltas (from 1 to 5) are 3.5, 3.3, 3.2, 3.1, 3.1, respectively. Compared to CV-MSER+, the speed-ups of Fast MSER V2 w.r.t. different deltas (from 1 to 5) are 3.2, 3.2, 3.2, 3.1, 3.0, respectively. Thus, the larger delta, the smaller speed-up.

Build

Windows

OpenCV 3.41

We have included the head files and the static library of OpenCV 3.41 in basic_thrid_libs/lib_opencv/341. The dynamic library of OpenCV 3.41 is in bin/x64/vc11/opencv_world341.zip. You can unzip opencv_world341.zip to get opencv_world341.dll. Note that opencv_world341.dll must be placed under bin/x64/vc11/.

Visual Studio 2012 (Release configuration, X64 platform)

Open all_projects.sln and build it, then you can test the comparison MSER algorithms.

MacOS

OpenCV 3

Please do not use OpenCV 4.

CLion

Please use CMakeLists.txt.

FAQ

If you have any questions about how to build this project, please tell me in 'Issues'. Thank you very much!