Does this algorithm support data with two channels (red and green channels)? Thanks!

Hi,

I think it is great to have the progress printed to the command line, but it results in a lot of text, and sometimes I want to see messages printed before I started image registration.

I have some few pieces of code that I use in my scripts to update the the textstring on the command line, instead of printing a new one:

str=[num2str(n), 'out of ', num2str(nFrames), 'frames registered'];
refreshdisp(str, prevstr, n);
prevstr=str; 

where refreshdisp looks like this:

function refreshdisp(str,prevstr,iteration)

if ~exist('iteration')
    iteration=2;
end

if iteration==1
    fprintf(str)
else
    fprintf(char(8*ones(1,length(prevstr))));
    fprintf(str);
end

also need to initialize prevstr.

Maybe this could be added to normcorre?

My videos have non-square pixels. In some (but not most) of my videos exit with the following error:

Error using normcorre_batch (line 283)
Index exceeds array bounds.

Here is the code I use to call normcorre_batch:

d1 = size(M1,1);
d2 = size(M1,2);

options_nonrigid = NoRMCorreSetParms(...
    'd1',d1,...
    'd2',d2,...
    ...
    'grid_size',[d1/4,d2/4],...
    'overlap_pre',[d1/16,d2/16],...
    'min_patch_size',[d1/16,d2/16],...
    'min_diff',[d1/32,d2/32,5],...
    'us_fac',50*d2/d1,...    
    'mot_uf',[2048/d1 2048/d2],...
    'max_dev',[d1/(170+2/3) d2/(170+2/3)],...
    'overlap_post',[d1/16 d2/16],...
    'max_shift',[d1/(512/70) d2/(512/70)],...
    ...    
    'bin_width',200,...        
    'init_batch',600,...
    'boundary',NaN);
tic; [M2,shifts2,template2,options_nonrigid] = normcorre_batch(M1,options_nonrigid); toc

A video that crashes normcorre can be found here (https://bit.ly/3hwT2k9).

line 35-38 in function normcorre.m

elseif strcmpi(ext,'mat')
filetype = 'mem';
Y = matfile(Y,'Writable',true);
sizY = size(Y);

This would not return the size of the data, it should be sizY = size(Y.Varname)

I have two image stacks. The size of them are 20482048600. I can get good results if I register each stack separately. However, if I register stack1 firstly and then use the average image of stack1(after registration) to register stack2. The alignment is pretty off.
Here's the average image of my stack1:

Here's the average image of my stack 2 after registration using the average of stack1:

I can get good results if I don't set template to register stack2.

I only used rigid motion correction. Here're my parameters:
NoRMCorreSetParms('d1',size(Y,1),'d2',size(Y,2),'bin_width',round(T),'max_shift',50,'us_fac',50,'init_batch',200);

Can this program combine multiple tiff files and then do the motion correction, and save the motion corrected tiff files?

I am using the correct_bidirectional_offset to correct data after resonance scanning. We just found out that when using [col_shift, M] = correct_bidirectional_offset(Y, nFrames, us) the pixel values of the output is different than the pixel values of the input.

Whereas the pixel values for the input varies between 31040 and 65534, the pixel values of the output varies between 26849 and 71852.

I usually run the code with an upsampling factor of 10 (default). I tried running with the us variabel set to 1 and then I got the expected minimum and maximum values on the output.

Hi Eftychios

When using the options.shifts_method = 'cubic', the imwarp function by default set to zero all values outside the image. So options.boundary does not have an effect on this, however you can fix that if you add the imwarp(*, *, 'FillValues', add_value).

one could easily fix this on apply_shifts.m by adding:
add_value = 0;
if strcmpi(options.boundary, 'nan')
add_value = nan;
end
on line: 213
and editing imwarp to accept 'FillValues' as add_value.
This would only work if you want those values to be 0 or nan, but dont quite know how to get the options.boundary options: 'copy' and 'template' implemented.

Diego

When using the cubic shift_method, I get this error:

Error using imwarp>checkImageAgreementWithTform (line 269)
The number of dimensions of the input image A must be 3 when the specified geometric transformation is 3-D.

for dm = 1:3; shifts_up(:,:,:,dm) = imwarp(shifts_temp(:,:,:,dm),tform,'OutputView',imref3d([options.d1,options.d2,options.d3])); end

• the size of shifts_temp(:,:,:,dm) is actually 1, but tform is three-dimensional.

Also there is a bug in line 306:

• buffer{i,j,k}(:,:,ind) = Mt2{ii};

• if nd == 2; buffer{i,j,k}(:,:,ind) = Mt2{ii}; end
• if nd == 3; buffer{i,j,k}(:,:,:,ind) = Mt2{ii}; end

Hello,
I am trying to understand this codebase. Can you please provide proper references and documentation for understanding this codebase better ?
For example, the code in correct_bidirectional_offset.m corrects phase artifacts.I'd like to understand the math behind this.It will be helpful to have the appropriate papers and research articles behind this code.
Thank you

Hey all,

I am planning on starting a python implementation of this package here. Let me know if you would like to contribute!

While registering each image to the template, normcorre_batch.m calls read_file.m to read in a number of frames equal to bin_width, which defaults to 10 and which we frequently set to be much larger. Every time read_file.m is called (at least for TIFFs) it must first read all the IFDs, which can take around 1 minute each time for large TIFFs. If we've read the code correctly, this means for a 30,000-frame movie with bin_width set to 300, the IFDs are read 100X... normcorre would spend 100 minutes on unnecessary file I/O. It might be good to revise the related code to prevent these re-reads.

Thanks.

Hi,
I started to encounter an error (see below) lately, and I wonder if you have any ideas why it might happen.

I am aligning an array of 5000 images, and this is the options I am using:

                Y1 = double(tmp_imArray);
                options_nonrigid = NoRMCorreSetParms('d1',size(Y1,1),'d2',size(Y1,2),...
                                       'grid_size',[64, 64],'mot_uf',4,'bin_width',43,...
                                       'max_shift',15,'max_dev',30,'us_fac',50);

                [tmp_imArray, ~, ~] = normcorre_batch(Y1, options_nonrigid);

This is the error:

Error using remove_boundaries (line 30)
Index exceeds matrix dimensions.

Error in shift_reconstruct (line 48)
    I = remove_boundaries(I,shifts,method,add_value);

Error in normcorre_batch>(parfor body) (line 332)
                                    M_fin{i,j,k} =
                                    shift_reconstruct(I_temp,shifts_up(i,j,k,:),diff_up(i,j,k),us_fac,Nr{i,j,k},Nc{i,j,k},Np{i,j,k},options.boundary,add_value);
                                    
Error in normcorre_batch (line 262)
        parfor ii = 1:lY

i am sorry if this is a dumb question but i can't figure out how to save the non rigid corrected tiff stack.

Hi, I've been using NoRMCorre to register single-plane 2-photon calcium imaging movies. Often, after performing non-rigid or rigid registration, the output movie contains one or more bright artifacts that do not correspond to any labeled cell structure. These artifacts usually consist of a bright 2x2 pixel square with more bright spots radiating out from this central spot in all 4 directions (see in lower right of attached image). I am running normcorre on 256x256 movies with 400+ frames, with the following parameters, all others being default:

options_nonrigid = NoRMCorreSetParms('d1',size(Y,1),'d2',size(Y,2),'grid_size',[64,64],'overlap_pre',[16,16],'mot_uf',4,'max_shift',[20,20],'max_dev',[8,8],'shifts_method','FFT');

I have tried changing patch and overlap size slightly, but the artifact remains. Using cubic or linear shifts reduces the intensity of the artifact, but it still remains. Any ideas of parameters to change to prevent this?

Hi, @epnev

I applied this NoRMCorre at a summer school, where we have access to a zebrafish data. The fish tail moves in a way such that partial field of view is compressing. I tried different ways to create patches (either set it to 4x3, or 8x8) for non-rigid correction in NoRMCorre, but the video looks almost the same between raw video and the registered video.. All the other parameters are used defaults as provided by demo.

Here I am attaching Google Slides where the video can be viewed: https://docs.google.com/presentation/d/1-BgCStw3yRDxCDEbarVGeDKUJShkI6FrXF06cKek0No/edit?usp=sharing

I don't understand why registered images look almost the same as the raw images. Should I set a lot more patches? Thanks a lot.

Hello,

We are using NormCorre and are checking our results on different computers. Each time we use a different computer with the same exact data set we are getting different results. We have identified the issue as a problem with the corr function.
According to our debugging, the corr(X,Y) function gives very different results when you loop through each column of the matrix i.e. call the corr function for each column and calculate corr(X(:,i), Y) at each column of X, versus when you just call the corr function once on the entire X matrix and Y.

We believe that this problem is what is leading to the different results for the same data set. We also believe that this problem is present in MATLAB versions of 2017a and previous versions. 2017b and beyond seem to be okay.

Have you experienced this problem and are there any solutions? Or did you already know this about the corr function and specifically chose to call it once on the entire matrix instead of looping through?

Thank you very much for you help!

Ginny
Richard Axel Laboratory

Hello,
I was wondering if you have tried to test the upper limit of how much you can increase bin_widths?

I was able to do motion correction on a 512x512x80000 .mat file using bin_widths of 10,000. I usually run this on a 40-core processor with 500G RAM. It takes about 2-3 hours with 12 parallel workers, which is called in the beginning with the gcp command and I do not enable memory mapping and also point to the path of file (without loading input .mat file into memory).

In general, the output seems motion corrected, however my videos have significant slow z-drift throughout. When I attempted to use bin_widths of 30,000, I run into several problems:

1. If I monitor the progress of the script, it appears to call call parpool several times (it calls 12 in the beginning and in the middle it also calls again, sometimes twice more), however the process is extremely slow and towards the end only a few cores seems to have a >0% load, even though the script technically called for more.

2. If I left the process run overnight, I get the following error:

Offset 4.0e-01 pixels due to bidirectional scanning detected.
Registering the first 10000 frames just to obtain a good template......done.
Template initialization complete. Now registering all the frames with new template.
The file was saved successfully. Elapsed time : 1140.249 s.
30000 out of 82521 frames registered, iteration 1 out of 1..Starting parallel pool (parpool) using the 'local' profile ...
connected to 12 workers.
Starting parallel pool (parpool) using the 'local' profile ...

Error using parfeval (line 58)
Parallel pool failed to start with the following error. For more detailed information, validate the profile 'local' in the Cluster Profile Manager.

Error in normcorre_batch_even (line 255)
future_results(i) = parfeval(@register_frame, 2, Yt,fftTempMat,fftTemp,patches,options);

Error in twostep_moco_applyshifts_clusterMAT (line 102)
[~,shifts,template] = normcorre_batch_even(fullname_R,options_mc,template);

Caused by:
Error using parallel.internal.pool.InteractiveClient>iThrowWithCause (line 676)
Failed to initialize the interactive session.
Error using parallel.internal.pool.InteractiveClient>iThrowIfBadParallelJobStatus (line 790)
The interactive communicating job failed with no message.

Is it possible that there is no upperlimit to parpool being called and so it keeps calling it until it runs out and errors? Any tips would be greatly appreciated. Or if there is a better way to deal with drift that would be great too.

Thank you!

The definition of v by videoreader needs to be repeated after the first readFrame command, otherwise the inital frame is lost. This only seems to be the case for avi files.

In NoRMCorre,when I selected a local tif file(transformed by a bmp file),an error appeared on function imwarp:
错误使用 imwarp
需要的 tform 应为以下类型之一:

images.geotrans.internal.GeometricTransformation

但其类型是 double。Meaning that I use the imwrap incorrectly,the tform should be this type:images.geotrans.internal.GeometricTransformation.But actually it was double.

I don't konw how to solve this problem,so ask for help and reply.Thank you!

I was trying to run NoRMCorre with the filename of a tiff as the input argument Y.
[M1,shifts1,template1] = normcorre(name,options_rigid)

turns out it grabs the info for the tiff (lines 24-26 of normcorre)
tiffInfo = imfinfo(Y);
filetype = 'tif';
T = length(tiffInfo);
but did not specify the variable sizY and ran into an error at line 49 of normcorre
nd = length(sizY)-1; % determine whether imaging is 2d or 3d

Thanks for a great package! Works great, and quicker than anything else I tried.

Would it be straight forward to apply shifts from one set of images onto another, e.g if we do dual color imaging?

I am running the following code:
`name = '/Users/eivinhen/Desktop/imreg_test_reg.tif';
Y = name;
iminfo = imfinfo(Y);

options_nonrigid = NoRMCorreSetParms('d1',iminfo(1).Height,'d2',iminfo(1).Width,'grid_size',[64,64],'mot_uf',4,'bin_width',50,'max_shift',25,'max_dev',20);
[M2,shifts2,template2] = normcorre_batch(Y,options_nonrigid);`

And it produces an error:
`Subscripted assignment dimension mismatch.

Error in cell2mat_ov (line 23)
X(xx_s(i):xx_f(i),yy_s(j):yy_f(j),zz_s(k):zz_f(k)) = ...

Error in normcorre_batch (line 306)
Mf{ii} =
cell2mat_ov(M_fin,xx_us,xx_uf,yy_us,yy_uf,zz_us,zz_uf,overlap_post,sizY)
- add_value;

Error in demo (line 46)
[M2,shifts2,template2] = normcorre_batch(Y,options_nonrigid);`

For example, I provide a template for the normcorre function, and this template has a different crop than the stack that I want to align. Will this cause a poor alignment or alignment artifacts, or does the method ignore edges like that?

Hi there. We recently came across an issue where NoRMCorre corrupts the output data in odd ways. We have used the library successfully for a long time, and experience the problem only with data from a new 2-photon microscope (Scanimage). The issue appears in some recordings, but not all. We are wondering if this is a known issue, and in any case if there is a fix?

Issue:

• Average image before NoRMCorre
• Average image after
• Raw data
• Parameters

Tried:

• Shifting values to above 0, and ensuring format is uint16
• Removing extreme values (helps, but zig-zag artefact is still present)

Vh Knut Åbjørsbråten

I was aligning a stack containing 701 frames using a bin width of 50 and got an error on alignment of the last bin:

Error using mat2cell (line 80)
Number of input vector arguments, 3, does not match the input matrix's number of dimensions, 2.

Error in normcorre_batch (line 266)
if nd == 2; Ytc = mat2cell(Ytm,d1,d2,ones(1,size(Ytm,ndims(Ytm)))); end

Error in imregSession (line 131)
[M1, shifts, ~] = normcorre_batch(Y1, options_nonrigid);

Error in runPipelineTMP (line 152)
imregSession(sessionID)

I ran it again using bin width of 40 and then it worked fine. Have not reproduced it with other combinations of stack sizes/bin widths.

I am looking to adopt NoRMCorre for a 4 channel dataset consisting of 3d stacks of images obtained along the axial plane of mice brain.So, the Field of View along z changes (thickness of brain slice=1 micron). Can i use NoRMCorre to correct motion artifacts in each image and correct alignment issues in every multichannel 3d stack ?
The dataset is 1024*1024 and z=44. We are imaging dendrites and looking to reconstruct the 3d view of dendrites from the 2d slices. There are no neurons.
Is NoRMCorre useful for my case?
Attached is a sample 3d stack of one of the channels.