2D matrix resize or interpolation

Question

Russell 2025-1-10

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2172834-2d-matrix-resize-or-interpolation

评论： Cris LaPierre 2025-1-27

SampleFiles.zip

Hello,

I'm new to matlab and I'm not sure what would be the best way to do the following.

I have an image/data matrix, I, of size (rows,cols) = (y,x) = [384,512].

The spatial resolution in the x direction is 0.078021 cm (total span is ~40cm)

The spatial resolution in the y direction is 0.078021 (total span is ~30cm)

Q1.

If I would like to have a finer sampling of the data by doubling the samples (eg Inew = [768, 1024], which function would you suggest?

Q2.

If I had another separate image but with offset x,y values, how would I interpolate this data onto the same grid as the other matrix?

Thank you for your help,

RR

ps. I found some info on the interp2 function? not sure if this is the best function to use?

ps. I have attached a zip file with

dicom image (of a radiotherapy dose on a single plane - my image)
excel workbook with the data extracted
an m script with what I tried to do but I got an error

Error using griddedInterpolant

The grid must be created from grid vectors which are strictly monotonically increasing.

Error in interp2>makegriddedinterp (line 228)

F = griddedInterpolant(varargin{:});

Error in interp2 (line 136)

F = makegriddedinterp(X, Y, V, method,extrap);

Error in LoadDosePlaneBailey2D (line 70)

Id_cgy_q = interp2(X,Y,Id_cgy,Xq,Yq);

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Cris LaPierre 2025-1-10

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2172834-2d-matrix-resize-or-interpolation#answer_1557292

编辑：Cris LaPierre 2025-1-10

Q1: B = imresize(A,[numrows numcols])

Q2: What type of images are you working with? Please share an example. You can attach images to your post using the paperclip icon.

16 个评论
显示 14更早的评论隐藏 14更早的评论

Cris LaPierre 2025-1-10

在 MATLAB Online 中打开

SampleFiles.zip

It looks like the pixel info and spreadsheet data are the same.

unzip('SampleFiles.zip')

% Use medical imaging toolbox

MI = medicalImage('ExampleDicomFileDosePlane.dcm')

MI =

medicalImage with properties: Pixels: [384x512 uint32] Colormap: [] SpatialUnits: "mm" FrameTime: [] NumFrames: 1 PixelSpacing: [0.7802 0.7793] Modality: 'RTDOSE' WindowCenter: [] WindowWidth: []

img = extractFrame(MI,1);

imshow(img,[])

% Resize

B1 = imresize(img,[798,1024]);

figure

imshow(B1,[])

Neither of these approaches take into consideration the actual pixel spacing. Instead, row/col index number is used. To display an image where the X and Y spacing is different, use the XData and YData inputs to imshow.

figure

ext = (size(img)-1).*MI.PixelSpacing

ext = 1×2

298.8198 398.2207

imshow(B1,[],"XData",[0 ext(2)],"YData",[0 ext(1)])

It's subtle, but you can compare the difference using imshowpair

RA = imref2d(size(img));

RB = imref2d(size(img),MI.PixelSpacing(2),MI.PixelSpacing(1));

imshowpair(img,RA,img,RB,'montage')

You could just as easily use the spreadsheet values. However, you lose the metadata information, which captures the pixel spacing.

dose = readmatrix('ExampleTxtDataOfDosePlane.xlsx');

imshow(dose,[])

B2 = imresize(dose,[768, 1024]);

figure

imshow(B2,[],"XData",[0 ext(2)],"YData",[0 ext(1)])

Cris LaPierre 2025-1-26

编辑：Cris LaPierre 2025-1-27

I'm defnitely not an expert in this space, so I'm still confused. Based on the metadata in the dsf files

Image_measured_30x40.dxf is actually 21.5x28.7
Image_prediction1_43x43.dxf is actually 31.1x31.1
Image_prediction2_30x40.dxf is actually 21.4x28.5

It looks like you have a 2-step process - first to extract the desired region using indexing, then resizing. Given that, I think the best approachis to use imresize on the extracted iage. Pick the approprite method for your task.

Cris LaPierre 2025-1-27

在 MATLAB Online 中打开

Image_prediction1_43x43.zip

Heres a function that will read in your dxf files. Not necessarily the best code, but does the job.

unzip Image_prediction1_43x43.zip

params = readDXF('Image_prediction1_43x43.dxf');

figure

imagesc(params.Pixels)

Xsz = params.Size1*params.Res1

Xsz = 311.4286

Ysz = params.Size2*params.Res2

Ysz = 311.4286

function data = readDXF(fname)

data = struct;

fid = fopen(fname);

while ~feof(fid)

ln = fgetl(fid);

[param,val] = strtok(ln,"=");

switch param

case {'[General]','[Geometry]','[Interpretation]','[Patient]','[Field]','[PortalDose]'}

continue

case '[Data]'

fseek(fid,0,'eof');

fgetl(fid);

otherwise

val = strtok(val,'=');

pat = asManyOfPattern(characterListPattern("0123456789-+."));

nums = extract(val,pat);

if length(val)==length(nums{1})

data.(param) = str2double(val);

else

data.(param) = val;

end

fclose(fid);

val = readmatrix(fname,'FileType','text','NumHeaderLines',48);

data.Pixels = val;

end

请先登录，再进行评论。

Answer 2

Les Beckham 2025-1-10

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2172834-2d-matrix-resize-or-interpolation#answer_1557293

For Q1, I would suggest imresize. Can you clarify what you mean by "offset x,y values" in Q2?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Russell 2025-1-10

Hello, I added some information as requested by Cris. Not sure if this helped?

The first image will be a dose plane

X size = ~40cm with 512pix samples at a res of ~0.078cm

Y size = ~30cm with 384pix samples at res of ~0.078cm

The second image will be a x-ray image but the detector may not exactly be centered (this is where the offset comes from)

xoffset, yoffset

I would like to resample either the dose plane calc on the same positions as the imager or vice versa.

Does this help clarify?

请先登录，再进行评论。

Answer 3

Walter Roberson 2025-1-10

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2172834-2d-matrix-resize-or-interpolation#answer_1557300

If I would like to have a finer sampling of the data by doubling the samples (eg Inew = [768, 1024], which function would you suggest?

If you were willing to have Inew = [767, 1023] instead of [768, 1024] then you can average adjacent rows and adjacent columns and manually insert the averaged values into the proper place. To be honest, though, calling interp2() is a heck of a lot easier.

The reason the output would be Inew = [767, 1023] instead of [768, 1024] is that you would be calculating new points between each existing point. If you had 3 points across then you have 2 intermediate points, for a total of 5. Calculating intermediate points gets you output of size 2*N-1 not output of size 2*N

Getting output exactly twice the original size requires interpolation at "just less than half" apart, like positions 1, 1.49, 1.98, 2.47,...

9 个评论
显示 7更早的评论隐藏 7更早的评论

Image Analyst 2025-1-10

在 MATLAB Online 中打开

If you want the same pixel values in the two pixels, then use the 'nearest' option of imresize. If you don't specify then I believe it uses bicubic interpolation.

img = [1,2,3,4;5,6,7,8]
img = 2×4
     1     2     3     4
     5     6     7     8
<mw-icon class=""></mw-icon>
<mw-icon class=""></mw-icon>
img2 = imresize(img, 2, 'nearest')
img2 = 4×8
     1     1     2     2     3     3     4     4
     1     1     2     2     3     3     4     4
     5     5     6     6     7     7     8     8
     5     5     6     6     7     7     8     8
<mw-icon class=""></mw-icon>
<mw-icon class=""></mw-icon>
img3 = imresize(img, 2)
img3 = 4×8
    0.5312    0.8047    1.3516    1.8750    2.3750    2.8984    3.4453    3.7188
    1.7188    1.9922    2.5391    3.0625    3.5625    4.0859    4.6328    4.9062
    4.0938    4.3672    4.9141    5.4375    5.9375    6.4609    7.0078    7.2812
    5.2812    5.5547    6.1016    6.6250    7.1250    7.6484    8.1953    8.4688
<mw-icon class=""></mw-icon>
<mw-icon class=""></mw-icon>
●

Note how it's handling the edge effects -- essentially the pixel value in your top row (1 in my example) is considered in the middle of the upsampled version (your row 2) so that's why in the bicubic version the leftmost is not exactly 1.

Russell 2025-1-14

Thanks. I guess I'm a little confusd over the spatial vector? I have something like this now

so I'm not sure how to get the interpolated image Iq from I, x[i], y[j] if I have another set of positions xq[i], yq[j] that I want to sample?

Russell 2025-1-15

matlab resize of dose plane image.pdf

Hi Walter,

I guess I'm still having problems understanding how to do this. I added another pdf showing what I'm trying to do. I'm unable to upload the calculated dicom planar dose image. I was wondering if what I'm asking is possible with what you showed me earlier? (this will be compared to the actual dose image measured that I showed earlier)

Thanks for your help.

请先登录，再进行评论。