3D-Imaging - Science method

There is no specific software that can directly convert a 2D image of the Köppen climate classification map into a 3D representation by simply inputting the image. Converting a 2D image to a 3D representation typically requires manual or automated reconstruction using specialized software or 3D modeling tools.

However, there are general-purpose 3D modeling and visualization software that can help you create a 3D representation of the Köppen climate classification map based on the information in the image. Some popular 3D modeling software options include:

Keep in mind that creating an accurate and detailed 3D representation of the Köppen climate classification map may require additional data and information beyond the 2D image. It might involve interpreting the climate zones, understanding the topography, and incorporating additional geospatial data to create a more realistic 3D model.

Overall, the process of converting a 2D image of the Köppen climate classification map into a 3D representation typically involves manual reconstruction and modeling using specialized 3D software tools.

this is online free tool that might help you, regarding colors, topologies etc .. you can lay the image on the contours

hope this helps

Hi Vladimir Farber, I think one of the fields of Light Field Images missing to be mentioned is the "Quality Assessment of Light Field Images" when they are propagating through a communication channel. For your kind reference, one of the published papers in this field is :

"Exploiting saliency in quality assessment for light field images."

Mahyar Radak

These links might be helpful, have a look;

1. https://www.mathworks.com/videos/introduction-to-matlab-with-image-processing-toolbox-90409.html

2. https://www.mathworks.com/matlabcentral/answers/169745-how-to-generate-a-digital-hologram

If by a 3D image you mean a volumetric reconstruction and not a stereoscopic image, I'd say that choosing the right method will depend on the file format and the aim of the reconstruction. All the options mentioned earlier can work fine in datasets of reasonable size. However, if, for some reason, you are trying to reconstruct a large size dataset (dozens of GB). A technical note we recently presented about the use of point clouds to reconstruct gigaresolution medical datasets might give you an idea of how to do it:

-DOI: 10.1096/fasebj.2021.35.S1.04298

Dear Pradip Singha , any stiff material consisting of low atomic number elements are suitable.

Examples are PMMA or polycarbonate.

Patient tables for Xray CT are built as sandwich of carbon fibers tissues (plate) -foam-carbon fiber tissue.

For small weight sample a thin carbon fiber plate is sufficient.

You may have seen them already.

In the attached linke there are some picture of such plates:

https://www.phd-24.de/fasern/carbongewebe/?gclid=EAIaIQobChMIl4qnm-XX8AIViP93Ch2JuAlREAAYASAAEgLM_fD_BwE

Can we measure height through analyzing one SEM image?

simplest way to 3D volume is here, do not need much resources, random forest algorithm.

Waqas Tanveer Sorry if you select the most convineant software please share it to me convenient

thank you in advance

Thank you @H. Douglas Morris for your help!

Thank you. ImageJ is a good application for the scale bar.

Hello, first of all, I would like to congratulate you on your new start, secondly, you should study survey and review papers on pattern recognition that will show you the existing methods and how the process should be also the same thing goes to the 3D reconstruction you should study image recomposition and stereoscopic images. Best of luck

Nasseh Khodaie Hi there!

Thanks for sharing the blog post.

I have no problem with reconstructed images as I am familiar with a plenty of software packages that can handle the segmentation.

My main problem is when CT images are passed to me without initial reconstruction which should normally be done right after scanning!

I guess I need to figure out how to do this in Matlab :-p

Hi, this m file can be useful.

https://uk.mathworks.com/matlabcentral/fileexchange/19058-cooc3d

We developed a software ourselves for 3D visualization, calculation and simulation of porous media. However, It is commercial. For free software, I'd like to recommend FDMSS ( http://porenetwork.com/download/fdmss/ ) developed by Kirill M. Gerke et al. The relative paper is )".

This is an old thread but let me put these code pieces here, so anyone needs in the future could use it! If your mask file was produced with SPM's plugin Marsbar, then you should have an additional sphere information file in .mat format for each of your .nii mask file (if you created your masks using sphere definition). Then you can use below code piece in Matlab to get the xyz coordinates for each of voxels in your mask. Obviously, make sure that you have SPM and Marsbar on your Matlab path (usually Marsbar comes together with SPM as default, but just in case!)

sphereInfo = maroi('myMask.mat');

volumeInfo=spm_vol('myMask.nii');

xyzCoordinates = realpts(sphereInfo,volumeInfo);

If you only have nifti file of your mask then these two lines of SPM code will be enough to extract such info

volumeInfo=spm_vol('myMask.nii')

[intensityValues,xyzCoordinates ]=spm_read_vols(volumeInfo);

I hope this helps!

Yes, you can convert a 3D model to a series of cross-sectional images using 3D Slicer. For example, you can load the STL file, then in Data module right-click to convert it to segmentation, then right-click on the created segmentation to convert it to labelmap. Post further questions on using 3D Slicer to discourse.slicer.org.

Elissa Mhanna you may used 3D Fusion from kinect SDK, its can be modified then you can apply a multiview.

Dear Anastasios Mastrokolias,

Look the link, maybe useful.

Regards,

Shafagat

If your data have a very high slice spacing in 3rd dimension and you are getting 2d slices results, you need to integrate neighbouring slices, with a fixed spacing of 1 mm using a linear interpolation to deal with this problem.

mathematically you can use Radon transform but you need more images not only three

Download Slicer it’s free

You can read/see them in a number of image viewers such as ITK-SNAP, Slicer, ImageJ etc. Or, you can play with the data in python/matlab.

BIM (buiding information modeling) can be a solution:

-Revit

-Archicad

- VisualArq

For details

Hi Danielle, have you seen the real vessel tube like structure after ECs culture on matrigel? How long did you culture? My guess is that ECs might group each other resembling vessels, what do you think? I may be wrong, ECs sprouts might form vessel like structure, in fact it might not be a real vessel. If there is a real vessel formation, you can control angiogenesis either by reducing the ECs numbers or using low concentration of matrigel or reducing both ECs number and matrigel concentration. Please update your successful attempt on how you control the rapid angiogenesis on matrigel substrata. Thanks.

It worked!! Thank you for the recommendation. I went into lab and plated the Matrigel a day in advance and added media on the top. The next day I plated my cells and the images were much better and the cells stayed in the same plane.

you can use voxel counter plugin in image j. After thresholding the stack, you can get the number of voxels for all the images of the z-stack. Since the size of each voxel is known you can get the volume multiplying by them.

I think David is on the right path. Here is what i did years ago.

To create a distance (thickness of soft tissue) map on the skin of a horned lizard (two species in this picture link or attached file)

An isosurface was created of the soft tissue or the skin. Then an isosurface was created of the bone.

Points from each surface (skin and bone) were compared to find the shortest distance between both sets of points.

Once the shortest distance was found, this was the value (distance or thickness) of the soft tissue. Each point was assigned a thickness value, in my case only the soft tissue or skin set of points was assigned a thickness value.

This thickness can be assigned a color gradient in the case of the horned lizards very thin or no thickness (horns) was set to white and varying in greater thickness assigned to red, yellow, green, and the thickest was blue and purple.

Free Software used was IsoSurf from Cambridge to create the isosurface but ParaView or MatLab can do the same types of isosurface calculations.

To compare the points I first sorted both sets of points by x then by y then z to reduce the amount of comparisons and then wrote my own code to find the shortest distance.

I think Excel can also do the sorts and find MIN distance but i have not tried this. Hope this helps.

The chisquare attack is a simble and famous method to test the Robustness of the security system against the attacks. It depend on probability analysis of stego image after put the information inside it using LSB method then compare it with the probability analysis of original image to check the difference between them. If the difference near zero it means no information inside the image otherwise if near one it means the image held the information.

Best regard

Zeyad Safaa Younus Saffawi

Im not an image analyst but many of my colleagues swear by ANTs for registrations (see Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. 2011. A reproducible evaluation of ANTs similarity metric performance in brain image registration. NeuroImage 54(3):2033–2044 DOI 10.1016/j.neuroimage.2010.09.025. ) ; we used this approach for mouse brain images (See Wood, T.C., Simmons, C., Hurley, S.A., Vernon, A.C., Torres, J., Dell’Acqua, F., Williams, S.C.R., Cash, D., 2016. Whole-brain ex-vivo quantitative MRI of the cuprizone mouse model. PeerJ 4, e2632. doi:10.7717/peerj.2632). Finally, i believe it is not essential to extract the brain from the skull prior to registration, but if you wish you can try the program that my colleague modified from FSL-BET and posted here: https://www.nitrc.org/projects/rbet/

IR and Visible Imaging are both of them full part of Computer Vision field. All topics of Visible Stereo Imaging applies to IR Stereo Imaging (feature point/contour detection, feature tracking, 3D Reconstruction from image sequences and so on...) . Neverthless, some computer vision techniques have to be specifically tuned/adapted to address IR applications. 

Some sample references are available by following:

Every language have advantages and disadvantages. In the case of C++ and Mathlab, I can say that Matlab is some times more easy to start test some ideas, bu  when you want to make a real product or software, you translate you Mathlab test to a more powerful language, like C++. Mathlab is used mostly for fast prototype and when you are sure of your idea you can spend more time writing a more complex and efficient program. So, for scientific porpouses, Mathlab is a beatiful language as C++ too. If you work with image processing, there are some librarys for C++ that can help you a lot, like OpenCV, CImg (for handle images formats and operations), ImageMagic, and others.

I have been using 1) Matlab and the Hedrick digitizing tools package (i.e., DLTdv and easyWand) and 2) Argus- an open source program (http://argus.web.unc.edu/). You might consider one or both of these routes. 

What's the update ? @ Dr. Tatyana

Thanks for your help.

Use BRATS dataset..

He is indeed Aparna. We believe they would benefit, we would benefit, and users would benefit.

This will get you some of the way:

for a given single image "i" where you have imported the image into MATLAB (so it is 1024×1024):

i_a = zeros(size(i));

i_a(i>0) = 1; %find non-zero values of image, ie. the cell.

i_a = imfill(i_a,'holes'); %fill in holes in image_a, have a look at the imfill documentation.

BW = edge(i_a); %find the edges of image_a

imagesc(BW); %have a look at the edges

[y,x] = find(BW); %find the indices of the non-zero (ie edge) values of BW.

z = ones(size(x))*n; %create a vector of z coordinates for the single image, multiply by n, where n is the height from the surface.

You can then repeat this for each image, z1:z18 and concatenate the x, y, and z coordinate vectors.

I segmented the coronary arteries using the region growing method as proposed by the article attached with this Question. I generated the 3D point cloud by using Isosurface command in MATLAB. Attached image shows a portion of point cloud, in this image,

1. The grey region represents a portion of segmented coronary arterial tree.

2. The black sports, represents centerline points of the segmented coronary tree.

3. The red points are the points for which the following equation holds

                          nx(x - xc) + ny(y - yc) + nz(z - zc) <= T

If you have only a small quantity and you would like to keep using the scaffold for tissue engineering you could compare the polymer density with the scaffold density to calculate porosity. This fast, inexpensive method has its limits on accuracy, though when I compared it with nCT scans I got comparable numbers.  With nCT (and thick enough fibers) you could also calculate surface area etc.

DOI: 10.1039/c3tb20995d

Hi Stéphane,

Thank you very much for your answer.

Best Regards

For non contact 3D measurement, LIDAR is the most accurate. Dense LIDAR is very expensive, sparse (such as automative) is cheaper, but will return only a sparse set of points with true depth, in the scene, Other less accurate methods are structured light, time of flight and stereo/multiview camera 3D.

One of the best software packages for graphs (2D and 3D) is Origin (by Origin Lab). The software is quite expensive but may be something that you can get through your organization. 

Thank you so much for your answer. 

Depends on the operations you have to implement, thecamount of data reduction allowed for visualization and especially the specifications regarding processing/response time.

While massive RAM and fast processors never hurt, some clever 'strategy' often helps: regarding eg. MRT slices, there is no need to have more than 10..20 layers in RAM at any time if your reconstruction uses a clever approach to eg. match contours from layer to layer. Such an approach greatly reduces memory swapping-in / swapping-out, thus speeding up the whole process.

But - as already stated - it depends on the individual application.

Blender is a free 3D modeling software to make pictures like the above image, you can try that. There is also google sketchup (simpler to learn but basic tooling)

Dear Renu Bisht

According to my researches pH= 5.3-5.8 was the best range for removing Cadmium.

Regards

Dear

Abdul Gajoum

It might be more effective approach if you can detect the individual object in a CT image, you have verity of options from the perceptive of feature extraction  ranging from image texture to Tensor flow pixel base image interpretation in terms of quantification. some of the examples are given bellow.

I would recommend you to contribute approach for image p reprocessing which might detect the sub-objects from the set of major image objects than use some machine learning techniques for classification as per nature of your problem statement.

Best Regards,

JAMIL AHMED

How do I do supervised learning for image recognition of clouds in photos using Python? - ResearchGate. Available from: https://www.researchgate.net/post/How_do_I_do_supervised_learning_for_image_recognition_of_clouds_in_photos_using_Python?tpr_view=wZvxZyTjRJf1NMU6LGacWvY13zsjnV2y1fkb_3#5823f37deeae39ae941dccf1 [accessed Nov 10, 2016].

Hi Liam

A UAV- (a.k.a. drone-) mounted LIDAR system is indeed a very suitable and well established tool to monitor artificial components such as pipeline infrastructure, construction frames and powerlines. As these structures are often very smooth and featureless, highly reflective, and consisting of very narrow components, UAV-photogrammetry often fails at accurately modeling these structures in 3D, due to lack of resolution and/or flawed tie-point matching on these surface types. E.g. for an accurate UAV-based 3D reconstruction of power lines and pylons including insulators, LIDAR is really the only option.

There are several LIDAR units available that have been sufficiently miniaturized for UAVs, e.g. from YellowScan, Velodyne and Riegl. They differ in scanning rates, range and processing options, related to their size and weight, which determines their use on rotary or fixed wing UAVs. The difficulty with UAV-borne LIDAR is that it requires an accurate trajectory reconstruction of the LIDAR system, i.e. highly accurate positioning and orientation information of the LIDAR unit along the entire flight path. Otherwise, it is impossible to align the scan strips to each other and georeference the block. This requires either (preferably) a very precise dual frequency GNSS antenna + receiver together with calibrated IMU unit, or a coupled photogrammetric system from which adjusted (calibrated) positions and orientations are calculated to then adjust the LIDAR block. 

Below are some references showing the process in detail: 

Wallace et al. (2012) Development of a UAV-LiDAR system with application to forest inventory. Remote Sensing 4, 1519-1543, doi:10.3390/rs4061519.

Jozkow et al. (2016) UAS topographic mapping with Velodyne LiDAR sensor. XXIII ISPRS congress, July 2016, Prague (Czech Republic). ISPRS Annals Volume III-1.

Maybe the AnalyzeSkeleton plugin for imageJ could help. It will make a binary version of your image based on a threshold, and skeletonize (removing pixels from both sides of a blob or thick line until a one-pixel-thick "skeleton" remains) its structures. For fluorescence images you could even split color channels and determine thresholds separately. The plugin can, then, count nodes (number of junctions) and measure length and thickness of branches. Junction density can be computed as a function of viewport area, which is usually provided by the microscopy software you are using or can be extrapolated from a scale bar. 

The official documentation for AnalyzeSkeleton:

http://imagej.net/AnalyzeSkeleton 

Hope it helps. Cheers!

Lindsey,

  I may be misreading your question, but it's important to note that not all 3D images are holograms- some are simply stacks of intensity images without phase data.  An example would be confocal microscopy 'image stacks'.

r/

David 

there is so many different techniques in optics for study of biological samples. I think digital holographic microscopy might be a good option for you because you can study dynamic and online processes with an accuracy of a few nanometer more and less...

Dear Aparna,

Thanks very much! I will try that later.

Best Regards

Liu

Hi Oktar,

could you share any link or document regarding the same.

Dear Ravi,

you said that you are new to confocal microscopy. This is why I want to say that the distortion you observe, especially in the Z-dimension, is a "normal" optical effect. It needs some reading in basic optics of a confocal laser scanning microscope. You may simply start (actually I did that, too) with reading Wikipedia articles about confocal microscopy and the point spread function (PSF). If you undertand the limits of your system you know what you can try to get better images.

But what can you do to get better Z-resolution? To be honest from the beginning: You will not get a perfect sphere-like representation of the bead in Z. There are ways to improve, but with a standard confocal there are limits.

1) Reduce the pinhole size. Usually the system is set to a pinhole size of 1 Airy Unit (AU). This decreases the volume you detect the emitted light from. You make the optical section more slim (in Z). It reduces the overall intensity but increases the signal to noise ratio. But at some point there is no more beneficial effect, I wouldn't go below 0.25 AU.

2) If you know the optical section thickness, then set the step size of your z-stack at least to a value that is half of it. An example: You have an optical section thickness of 980nm, your step sze should be 490nm or smaller. Keep in mind, if you reduce pinhole size the optical section gets smaller and you will need to reduce the step size as well. This will not kill your fluorescent bead but live samples do not like too much laser power.

3) As James said, check if you use the best combination of refractive index values (RI). It might be better to use a 40x oil objetive on some fixed sample (with an embedding medium that resembles the RI of oil) than a 63x water objective.

4) There is the possibility to do post-processing of confocal z-stacks, called deconvolution. But this is very tricky to do and requires even more image setting adjustments.

A last thing to the dimensions in standard confocal microscopy: the size of an object you see/measure in an unprocessed confocal image is never correct. They appear always bigger than they really are. Especially when talking about subcellular compartments etc.

Hope this could help you a little bit.

Best,

Falco

Hello,

You might find mesh comparison tools, such as PolyMeCo or Meshlab, useful. You can take a look at:

Meshlab: http://meshlab.sourceforge.net/

PolyMeCo: http://sweet.ua.pt/sss/projects/PolyMeCo/index.html

Cheers!

Dears

may this link support your subject

Regards

If you know the histone centers, you know the DNA segments wrapped around them. 

There are some X-ray crystal structures of individual nucleosomes (e.g., Luger et al. (1997) Nature 389: 251-260) . . .

. . . but the overall conformation of the promoter (i.e., the relative orientation of nucleosomes) will be modulated by intra- and inter-nucleosome interactions, promoter-binding proteins and sequentially remote but spatially close chromatin region(s).

You can build a coarse-grained model of the entire promoter (based on the structural and functional experimental data and models) . . .

. . . let it relax in MD simulations and generate an ensemble of structures but it is more likely that you will get a NASA-style artist's impression rather than a realistic model.

Shareware has a list

Perhaps there is some useful information in this discussion:

simple convert them raster to ascii in arc gis software. Read documentation first

Asymptote is based on LaTeX but optimized with C++ to draw efficiently, especially in 3D. I recommend you to learn it. Other options in LaTeX are Tikz, Tikz-euclid packages!

Hi Chiranjib,

please check the following paper from our group:

the method consists of GUV dispersion in a fluid (warm) agarose gel that gelifies at room temp. It is well suited for GUVs but also small vesicles (LUVs).

I hope this might help you.

Best regards,

Rafael

Thanks Narasim. I like the idea of converting data to pixels but that's too general. I will most probably use the hiveplot. 

Please refer to Antonio Messina's answer!

You can find more about FDM technique in attached paper.

Hi,

When doing coregistration the transformation is actually stored in the header of the source image.

If you want to reslice (interpolate) your source image to also have the same image dimensions and orientation as your reference image, you should 'coregister&reslice'. Then a new image is generated (r+source image)

Hope this helps.

Kind regards,

Mathijs Raemaekers

There are pros and cons for either FSL or SPM. Personally, I prefer the latest SPM12 version. The new user manual is easily understandable, but you need to be aware that you need a Matlab license to run SPM!

Puhhh - this is just to much! Seriously, I recommend to seperate your problem into some parts and open new questions about it.

What I understood so far is, that you want to make a comparison of two (or more) different techniques with a meshed 3D model. In the field of photogrammetry a comparison of different 3D models is usually done like this:

I hope it helps!

Regards

Olaf

We have recently explored Chimera and Vaa3D.  Only tried 4D in Chimera which works well except for a Zdepth rendering problem when dealing with volumetric data.  Surfaces have no problem, but for volumes, one channel always renders in front of the others.

If you write the expanded version of the rendering equation for all pixels, could you optimize it for light source direction, so that the equation produces the given pixels?

We use a virtual light stage, but we have an estimation of geometry. Do you have or plan to estimate geometry? Or is it possible to make a hard assumption on surface normals? Is interaction allowed? 

Also, have you seen this: "Specularity removal in images and videos"; SP Mallick, T Zickler, PN Belhumeur; ECCV 2006

Hi Chiththaka,

another software which can do this really well is Imod (University of Colorado). There is an option to align series of images and also it allows you to do segmentation. You will just need to convert your tifs to mrc file, which is easy (command tif2mrc). 

Best wishes,

Petr

Hi. There are a lot of 3D softwares out there, you can try the students' versions of Autodesk softwares such as 3ds Max, Maya, Inventor or Autocad. Also Solidworks is a good idea. and if you want a free software, you can try Blender, there are multiple options. 

Generally one should develop/obtain a code to simulate one or multiphase flow in the network. Maybe the OpenPNM (as Philip said) might have some feature. Also, some researcher may share their codes with you upon request. My experience with Imperial College London (team of Prof. Blunt) was most of the times appealing. 

You can find some information about 4 different algorithms in the Paper "Change detection in Terrestrial Laser Scanner Data Via Point Cloud Correspondence" from M. Tsakiri & V. Anagnostopoulos.

The algorithms they compared are Nearest Neighbour, Nearest Neighbour with local modelling, Normal Shooting and ICP.