The sigma parameter of vnlbayes gives this description:
"add noise of standard deviation sigma"

Which means we have to pass the non noisy frames as input.

Please make it possible to give noisy frames and the estimated sigma.

Hey, will the algorithm performs better when an optical flow is computed?

Thank you for the source.

There are two variants of the VNLB,VNLB-S and VNLB-H, with patch sizes 7×7×2 and 10×10×2.And the default parameter of the code is VNLB-H with patch sizes 7×7×2.

l want to run VNLB-H with patch sizes 10×10×2.l tred to enter the command like this on the test video tennis ,but it didn't run.

vnlb-gt.sh /my/video/frame-%03d.png 1 150 10 out-folder -px1 10 -px2 10

So could you tell me how to run VNLB-H with patch sizes 10×10×2 and VNLB-S.

Hello,

Is there any benchmarking code to go along with this? Maybe the script used to run the experiments in the paper? I am getting a strangely low PSNR with the current code. I will keep testing, but I am unable to duplicate your impressive results.

Hello,

How would the codebase managers feel about creating a Python API? This December I may have some time to write a wrapper to call your library directly from Python using SWIG. Is this something that would be interesting? I can create a PR for your code, and it would look something like this repo.

To give some more context, this Python & SWIG & C++ combo is also how FAISS adds a Python API to their C++ code.

[ 66%] Linking CXX shared library ../../lib/libvnlb.so
/usr/bin/ld: cannot find CMakeFiles/vnlb.dir/LibMatrix.cpp.o: No such file or directory
collect2: error: ld returned 1 exit status
[ 66%] Built target vnlb
[ 72%] Building CXX object src/CMakeFiles/psnr.dir/psnr.cpp.o
[ 77%] Linking CXX executable ../bin/psnr
[ 77%] Built target psnr
[ 83%] Building CXX object src/CMakeFiles/awgn.dir/awgn.cpp.o
[ 88%] Linking CXX executable ../bin/awgn
[ 88%] Built target awgn
[ 94%] Building CXX object src/CMakeFiles/vnlbayes.dir/main_vnlb.cpp.o
make[2]: *** No rule to make target 'lib/libvnlb.so', needed by 'bin/vnlbayes'. Stop.
[100%] Built target vnlbayes

ls bin

awgn psnr tvl1flow tvl1flow-seq.sh vnlb-gt.sh vnlb.sh

Everything but vnlbayes is compiling. Can you suggest a workaround? Thanks!

When I run the main function on a sequence with 2 images. It always give us a warning "OpenBLAS Warning : Detect OpenMP Loop and this application may hang. Please rebuild the library with USE_OPENMP=1 option". Also it seems that the program does no stop even after 5 mins. How can I solve this issue?

Thank you very much for open source related code.When l complied the code ,there are two errors.l have tried several times,but l don't know how to solve it.

`shentu@shentu-Lenovo-Product:~/code/vnlb-master/build$ make
[ 11%] Built target iio

[ 22%] Built target tvl1flow

[ 33%] Built target tvl1flow-bin

[ 50%] Built target vidutils

[ 66%] Built target vnlb

[ 77%] Built target awgn

[ 83%] Linking CXX executable ../bin/vnlbayes

/usr/bin/ld: /usr/local/lib/liblapack.a(xerbla.o): undefined reference to symbol

'_gfortran_string_len_trim@@GFORTRAN_1.0'

//usr/lib/x86_64-linux-gnu/libgfortran.so.3: error adding symbols: DSO missing from command line

collect2: error: ld returned 1 exit status

src/CMakeFiles/vnlbayes.dir/build.make:101: recipe for target 'bin/vnlbayes' failed

make[2]: *** [bin/vnlbayes] Error 1

CMakeFiles/Makefile2:389: recipe for target 'src/CMakeFiles/vnlbayes.dir/all' failed

make[1]: *** [src/CMakeFiles/vnlbayes.dir/all] Error 2

Makefile:127: recipe for target 'all' failed

make: *** [all] Error 2
`

Hi,

The default patch size seems to be 7x7x2.

However in your paper "Video denoising via empirical Bayesian estimation of space-time patches", you compared 7x7x2 and 10x10x2 and shows that the latter has better performance.
I think this should be set the default.

Thank you so much for your excellent work.

But I have met an error every time when I run the command 'vnlbayes -i /my/video/frame-%03d.png -f 1 -l 10 -sigma 10'

terminate called after throwing an instance of 'std::runtime_error'
what(): VideoNLB::computeBayesEstimate: insufficient patches for given rank/patch dimension.
Aborted (core dumped)

Have you ever met it before or what can I do to solve this problem?

Thank you so much.