Correct image for perspective distortion
Locked
Correct image for perspective distortion
 
|
Hi
I have some 2D photographic images that I want to correct for perspective distortion. I placed a square frame around the object I was photographing to use for correction. I can select the area within the frame in ImageJ. Is it possible to then stretch out the narrower part of that selection to form the square? Thanks Kathy -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of IMAGEJ automatic digest system Sent: 25 July 2013 05:00 To: [hidden email] Subject: IMAGEJ Digest - 23 Jul 2013 to 24 Jul 2013 (#2013-216) There are 7 messages totaling 2640 lines in this issue. Topics of the day: 1. newbie needs to know about 3D (4) 2. Announcement: new ImageJ-based textbook ("Advanced Methods") 3. Intensity analysis of a ROI from 2 channels 4. Measuring corrosion pit volumes. -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ---------------------------------------------------------------------- Date: Tue, 23 Jul 2013 23:56:15 -0400 From: "[hidden email]" <[hidden email]> Subject: newbie needs to know about 3D HOW TO MAKE A 3-D RENDERING OF OUTLINES OF NUCLEARBOUNDARIES IN AN EVENLY SPACED SET OF STAINED SECTION THROUGH A MAMMALIANBRAIN. What we thought would be a reasonable approach: 1. 1. Scan or photograph each section in the series, 2. 2. Align the sections, ideally by reference toblockface photos taken during the sectioning process. Alterniatively, open thesections one at a time in photoshop, copying each section and placing it a alayer of a .psd file. By manipulating the images of two layers so as tonear-exactly place each layer in line with its predecessor. 3. 3, Then crop the .psd file so each image frame isthe same, and the positions of the sections in the frames define the boundariesof the eventual 3D rendering. 4. 4. Save each layer as a .tif file, and place all ofthem in a folder. 5. 5. Drop the folder into the ImageJ window so as tomake a stack. 6. 6. In each image in the stack, using the freehandtrace the outlines of the nucleus of interest. HERE THE WHOLE APPROACH COLLAPSES IN CHAOS, SINCE WHEN WE TRACE ANOUTLINE ON AN IMAGE, THEN SCROLL TO THE NEXT IMAGE IN THE STACK, THE OUTLINE LEAVES THE IMAGE AND COMES WITH THESCROLL. ATTEMPTING TO DRAW ANYTHING CAUSES THE OUTLINE TO VANISH AND ONE CANNOT DRAW AN OUTLINE AROUND THE NUCLEUS IN THE NEXT IMAGE. Where did we go wrong, and where can we find any instructions at allabout how to get the images and or outlines into Image J so as to continue withthe 3D rendering.? Any information will be greatly appreciated, John I. Johnson [hidden email]. -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ Date: Wed, 24 Jul 2013 10:49:41 +0200 From: José María Mateos <[hidden email]> Subject: Re: newbie needs to know about 3D 2013/7/24 [hidden email] <[hidden email]>: > HERE THE WHOLE APPROACH COLLAPSES IN CHAOS, SINCE WHEN WE TRACE ANOUTLINE ON AN IMAGE, THEN SCROLL TO THE NEXT IMAGE IN THE STACK, THE OUTLINE LEAVES THE IMAGE AND COMES WITH THESCROLL. ATTEMPTING TO DRAW ANYTHING CAUSES THE OUTLINE TO VANISH AND ONE CANNOT DRAW AN OUTLINE AROUND THE NUCLEUS IN THE NEXT IMAGE. I understand you would have to use the ROI manager (Analyze > Tools > ROI Manager...) to draw different selections on different slices of your stack. Best, José María Mateos. -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ Date: Wed, 24 Jul 2013 10:40:58 +0200 From: Stephan Saalfeld <[hidden email]> Subject: Re: newbie needs to know about 3D Hi John, please read about and check the plugin TrakEM2 http://fiji.sc/TrakEM2 which is included in Fiji. It's designed for exactly what you are describing, reconstructing 3D structures from section series, some learning curve involved. It also enables automatic section series alignment about which you can read here: http://fiji.sc/Elastic_Alignment_and_Montage AND PLEASE STOP SCREAMING!!!! ;) Cheers, Stephan On Tue, 2013-07-23 at 23:56 -0400, [hidden email] wrote: > HOW TO MAKE A 3-D RENDERING OF OUTLINES OF NUCLEARBOUNDARIES IN AN EVENLY SPACED SET OF STAINED SECTION THROUGH A MAMMALIANBRAIN. > > What we thought would be a reasonable approach: > > > 1. 1. Scan or photograph each section in the series, > > > 2. 2. Align the sections, ideally by reference toblockface photos taken during the sectioning process. Alterniatively, open thesections one at a time in photoshop, copying each section and placing it a alayer of a .psd file. By manipulating the images of two layers so as tonear-exactly place each layer in line with its predecessor. > > > 3. 3, Then crop the .psd file so each image frame isthe same, and the positions of the sections in the frames define the boundariesof the eventual 3D rendering. > > > 4. 4. Save each layer as a .tif file, and place all ofthem in a folder. > > > 5. 5. Drop the folder into the ImageJ window so as tomake a stack. > > > 6. 6. In each image in the stack, using the freehandtrace the outlines of the nucleus of interest. > > > HERE THE WHOLE APPROACH COLLAPSES IN CHAOS, SINCE WHEN WE TRACE ANOUTLINE ON AN IMAGE, THEN SCROLL TO THE NEXT IMAGE IN THE STACK, THE OUTLINE LEAVES THE IMAGE AND COMES WITH THESCROLL. ATTEMPTING TO DRAW ANYTHING CAUSES THE OUTLINE TO VANISH AND ONE CANNOT DRAW AN OUTLINE AROUND THE NUCLEUS IN THE NEXT IMAGE. > > Where did we go wrong, and where can we find any instructions at allabout how to get the images and or outlines into Image J so as to continue withthe 3D rendering.? > > Any information will be greatly appreciated, > > John I. Johnson > [hidden email]. > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ Date: Wed, 24 Jul 2013 11:03:03 +0200 From: Robert Kirmse <[hidden email]> Subject: Re: newbie needs to know about 3D Hi John, I don't know if I understood the imaging process of your sample correctly but here are my thought for your process. If you haven't already, download the Fiji package it is essentially ImageJ with a bunch of pre-installed Plugins, including a lot that deal with stack registration i.e. alignment and segmentation. Of course you can also download the plugins for ImageJ seperately Without having seen your images I would already start in ImageJ/Fiji, and import the whole series into a stack 1. Fiji > File > Import > Image Sequence (your files need to be numbered) 2. Fiji > Plugins > Registration > Stackreg (this should already help you with a decent alignment, but also try other methods from the registration plugins to see what is best for you data. This should give you a nicely aligned stack that you can movie through. Crop the stack to your needs in ImageJ/Fiji 3. Segmentation, You need a good Segmentation editor that can handle different objects that are traced either manually or automatically. In these editors you won't have the problem you described because the program takes care of the outlines you draw and to which object they belong. Afterwards you can render the outlines that belong to each object. My experience with Segmentation editors in ImageJ is limited so I can't give a good recommendation I normally used the imod package (developed for electron microscopy) for manual segmentation. However even with good tools manual segmentation is quite time intensive. Also if you "just" want to generate a 3D rendering try Fiji > Plugins > Volume Viewer Here you have to experiment with the display option and the transfer curves to bring out the feature you want and surpress the unwanted features. I guess for what you want the option "Volume" under Mode is probably what you want. You could also have a look at the Ilastik package in addition to the Segmentation plugins in Fiji which is a good trainable segmentation tool that also offers 3D rendering options. Anyway, Segmentation will require some learning time, so be prepared to spent some time researching and trying different tools. Cheers Robert -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ Date: Wed, 24 Jul 2013 11:56:00 +0200 From: Burger Wilhelm <[hidden email]> Subject: Announcement: new ImageJ-based textbook ("Advanced Methods") Dear ImageJ community, we are pleased to announce a new follow-up volume "Advanced Methods" of our ImageJ-based textbook series on "Principles of Digital Image Processing". This 3rd volume contains an in-depth, algorithmic treatment of several classic topics including - automatic thresholding, - filters for color Images, - edge detection in color images, - edge-preserving smoothing filters, - Fourier shape descriptors, - scale-invariant local features (SIFT). As usual, the corresponding Java/ImageJ source code (with many example plugins) is freely available to everbody, including a complete scale space and SIFT implementation that was newly developed from ground up. Note that we also restructured our source code by merging all core classes into a single JAR file and putting ImageJ plugins into corresponding Java packages. (Thanks to Wayne for making this possible!) For all details and downloads please visit http://imagingbook.com/books/englisch-edition-3-vol-softcover/ Due to technical difficulties, our web site had to be rehosted recently and we apologize for any inconveniences in this context. We hope that our newly designed site is stable and users find the material helpful. Please feel free to send comments and suggestions. With kind regards, Wilhelm Burger & Mark J. Burge www.imagingbook.com -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ Date: Wed, 24 Jul 2013 09:57:05 -0500 From: Dinesh Joshi <[hidden email]> Subject: Intensity analysis of a ROI from 2 channels Hell everyoneI want to measure the intensity of a region of interest (ROI) from 2 channels. In the attached image I would like to measure the intensity of DAB in the regions of Neurofilaments (alexa fluor 488 fluorescence). Is there any way to pick a ROI in the merge image and get the intensity of that ROI from both DAB and neurofilament ? Thanks Dinesh C Joshi -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ Date: Wed, 24 Jul 2013 18:01:48 +0000 From: "Bryan, Charles" <[hidden email]> Subject: Measuring corrosion pit volumes. Hi All, I am new to ImageJ, and am trying to figure out how to measure the volumes of corrosion pits on polished metal surfaces. The raw data consists of image stacks collected with a laser confocal microscope, which are readily read and interpreted by ImageJ. The surface is defined by the layer with the brightest intensity. I need to be able to correct of surface tilt, determine the elevation in the stack for each x,y point, interpolate between adjacent layers to fill in holes (sometimes the pits walls are too steep, and leave shadows in the pit), and then calculate the pit volume and area. I think I can work out the individual steps if I have to, but suspect that the macros have already been developed, since this seems like a standard type of analysis. Does anyone have any suggestions? Thanks, Charles Bryan [hidden email] -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html ------------------------------ End of IMAGEJ Digest - 23 Jul 2013 to 24 Jul 2013 (#2013-216) ************************************************************* -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Re: Correct image for perspective distortion
|
|
Hi Kathy,
yes. In Fiji, you will find the Plugin Plugins > Transform > Landmark Correspondences documented here: http://fiji.sc/Landmark_Correspondences Then, open your image, marks the four corners of the square with the multi-point selection tool, then make a new image into which you paint a square, and mark its corners with the multi-point selection tool (same order such that 1 corresponds to 1, 2 corresponds to 2, etc.). Now call the plugin and choose the distorted image as source, the square as template, and Perspective for the transformation class. OK will generate the undistorted image. Best, Stephan On Thu, 2013-07-25 at 10:05 +0000, Chandler, Kathy wrote: > Hi > > I have some 2D photographic images that I want to correct for perspective distortion. I placed a square frame around the object I was photographing to use for correction. I can select the area within the frame in ImageJ. Is it possible to then stretch out the narrower part of that selection to form the square? > > Thanks > > Kathy > > -----Original Message----- > From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of IMAGEJ automatic digest system > Sent: 25 July 2013 05:00 > To: [hidden email] > Subject: IMAGEJ Digest - 23 Jul 2013 to 24 Jul 2013 (#2013-216) > > There are 7 messages totaling 2640 lines in this issue. > > Topics of the day: > > 1. newbie needs to know about 3D (4) > 2. Announcement: new ImageJ-based textbook ("Advanced Methods") > 3. Intensity analysis of a ROI from 2 channels > 4. Measuring corrosion pit volumes. > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ---------------------------------------------------------------------- > > Date: Tue, 23 Jul 2013 23:56:15 -0400 > From: "[hidden email]" <[hidden email]> > Subject: newbie needs to know about 3D > > HOW TO MAKE A 3-D RENDERING OF OUTLINES OF NUCLEARBOUNDARIES IN AN EVENLY SPACED SET OF STAINED SECTION THROUGH A MAMMALIANBRAIN. > > What we thought would be a reasonable approach: > > > 1. 1. Scan or photograph each section in the series, > > > 2. 2. Align the sections, ideally by reference toblockface photos taken during the sectioning process. Alterniatively, open thesections one at a time in photoshop, copying each section and placing it a alayer of a .psd file. By manipulating the images of two layers so as tonear-exactly place each layer in line with its predecessor. > > > 3. 3, Then crop the .psd file so each image frame isthe same, and the positions of the sections in the frames define the boundariesof the eventual 3D rendering. > > > 4. 4. Save each layer as a .tif file, and place all ofthem in a folder. > > > 5. 5. Drop the folder into the ImageJ window so as tomake a stack. > > > 6. 6. In each image in the stack, using the freehandtrace the outlines of the nucleus of interest. > > > HERE THE WHOLE APPROACH COLLAPSES IN CHAOS, SINCE WHEN WE TRACE ANOUTLINE ON AN IMAGE, THEN SCROLL TO THE NEXT IMAGE IN THE STACK, THE OUTLINE LEAVES THE IMAGE AND COMES WITH THESCROLL. ATTEMPTING TO DRAW ANYTHING CAUSES THE OUTLINE TO VANISH AND ONE CANNOT DRAW AN OUTLINE AROUND THE NUCLEUS IN THE NEXT IMAGE. > > Where did we go wrong, and where can we find any instructions at allabout how to get the images and or outlines into Image J so as to continue withthe 3D rendering.? > > Any information will be greatly appreciated, > > John I. Johnson > [hidden email]. > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > Date: Wed, 24 Jul 2013 10:49:41 +0200 > From: José María Mateos <[hidden email]> > Subject: Re: newbie needs to know about 3D > > 2013/7/24 [hidden email] <[hidden email]>: > > HERE THE WHOLE APPROACH COLLAPSES IN CHAOS, SINCE WHEN WE TRACE ANOUTLINE ON AN IMAGE, THEN SCROLL TO THE NEXT IMAGE IN THE STACK, THE OUTLINE LEAVES THE IMAGE AND COMES WITH THESCROLL. ATTEMPTING TO DRAW ANYTHING CAUSES THE OUTLINE TO VANISH AND ONE CANNOT DRAW AN OUTLINE AROUND THE NUCLEUS IN THE NEXT IMAGE. > > I understand you would have to use the ROI manager (Analyze > Tools > ROI Manager...) to draw different selections on different slices of your stack. > > Best, > > José María Mateos. > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > Date: Wed, 24 Jul 2013 10:40:58 +0200 > From: Stephan Saalfeld <[hidden email]> > Subject: Re: newbie needs to know about 3D > > Hi John, > > please read about and check the plugin TrakEM2 > > http://fiji.sc/TrakEM2 > > which is included in Fiji. It's designed for exactly what you are describing, reconstructing 3D structures from section series, some learning curve involved. It also enables automatic section series alignment about which you can read here: > > http://fiji.sc/Elastic_Alignment_and_Montage > > AND PLEASE STOP SCREAMING!!!! ;) > > Cheers, > Stephan > > > > On Tue, 2013-07-23 at 23:56 -0400, [hidden email] wrote: > > HOW TO MAKE A 3-D RENDERING OF OUTLINES OF NUCLEARBOUNDARIES IN AN EVENLY SPACED SET OF STAINED SECTION THROUGH A MAMMALIANBRAIN. > > > > What we thought would be a reasonable approach: > > > > > > 1. 1. Scan or photograph each section in the series, > > > > > > 2. 2. Align the sections, ideally by reference toblockface photos taken during the sectioning process. Alterniatively, open thesections one at a time in photoshop, copying each section and placing it a alayer of a .psd file. By manipulating the images of two layers so as tonear-exactly place each layer in line with its predecessor. > > > > > > 3. 3, Then crop the .psd file so each image frame isthe same, and the positions of the sections in the frames define the boundariesof the eventual 3D rendering. > > > > > > 4. 4. Save each layer as a .tif file, and place all ofthem in a folder. > > > > > > 5. 5. Drop the folder into the ImageJ window so as tomake a stack. > > > > > > 6. 6. In each image in the stack, using the freehandtrace the outlines of the nucleus of interest. > > > > > > HERE THE WHOLE APPROACH COLLAPSES IN CHAOS, SINCE WHEN WE TRACE ANOUTLINE ON AN IMAGE, THEN SCROLL TO THE NEXT IMAGE IN THE STACK, THE OUTLINE LEAVES THE IMAGE AND COMES WITH THESCROLL. ATTEMPTING TO DRAW ANYTHING CAUSES THE OUTLINE TO VANISH AND ONE CANNOT DRAW AN OUTLINE AROUND THE NUCLEUS IN THE NEXT IMAGE. > > > > Where did we go wrong, and where can we find any instructions at allabout how to get the images and or outlines into Image J so as to continue withthe 3D rendering.? > > > > Any information will be greatly appreciated, > > > > John I. Johnson > > [hidden email]. > > > > -- > > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > Date: Wed, 24 Jul 2013 11:03:03 +0200 > From: Robert Kirmse <[hidden email]> > Subject: Re: newbie needs to know about 3D > > Hi John, > I don't know if I understood the imaging process of your sample correctly but here are my thought for your process. > > If you haven't already, download the Fiji package it is essentially ImageJ with a bunch of pre-installed Plugins, including a lot that deal with stack registration i.e. alignment and segmentation. Of course you can also download the plugins for ImageJ seperately > > Without having seen your images > I would already start in ImageJ/Fiji, > and import the whole series into a stack > > 1. Fiji > File > Import > Image Sequence (your files need to be numbered) 2. Fiji > Plugins > Registration > Stackreg (this should already help you with a decent alignment, but also try other methods from the registration plugins to see what is best for you data. > This should give you a nicely aligned stack that you can movie through. > Crop the stack to your needs in ImageJ/Fiji > > 3. Segmentation, > You need a good Segmentation editor that can handle different objects that are traced either manually or automatically. In these editors you won't have the problem you described because the program takes care of the outlines you draw and to which object they belong. > Afterwards you can render the outlines that belong to each object. > My experience with Segmentation editors in ImageJ is limited so I can't give a good recommendation I normally used the imod package (developed for electron microscopy) for manual segmentation. > However even with good tools manual segmentation is quite time intensive. > > Also if you "just" want to generate a 3D rendering try Fiji > Plugins > Volume Viewer Here you have to experiment with the display option and the transfer curves to bring out the feature you want and surpress the unwanted features. I guess for what you want the option "Volume" under Mode is probably what you want. > > You could also have a look at the Ilastik package in addition to the Segmentation plugins in Fiji which is a good trainable segmentation tool that also offers 3D rendering options. > > Anyway, Segmentation will require some learning time, so be prepared to spent some time researching and trying different tools. > > Cheers > Robert > > > > > > > > > > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > Date: Wed, 24 Jul 2013 11:56:00 +0200 > From: Burger Wilhelm <[hidden email]> > Subject: Announcement: new ImageJ-based textbook ("Advanced Methods") > > Dear ImageJ community, > > we are pleased to announce a new follow-up volume "Advanced Methods" of our ImageJ-based textbook series on "Principles of Digital Image Processing". This 3rd volume contains an in-depth, algorithmic treatment of several classic topics including > > - automatic thresholding, > - filters for color Images, > - edge detection in color images, > - edge-preserving smoothing filters, > - Fourier shape descriptors, > - scale-invariant local features (SIFT). > > As usual, the corresponding Java/ImageJ source code (with many example plugins) is freely available to everbody, including a complete scale space and SIFT implementation that was newly developed from ground up. Note that we also restructured our source code by merging all core classes into a single JAR file and putting ImageJ plugins into corresponding Java packages. (Thanks to Wayne for making this possible!) > > For all details and downloads please visit http://imagingbook.com/books/englisch-edition-3-vol-softcover/ > > Due to technical difficulties, our web site had to be rehosted recently and we apologize for any inconveniences in this context. We hope that our newly designed site is stable and users find the material helpful. Please feel free to send comments and suggestions. > > With kind regards, > > Wilhelm Burger & Mark J. Burge > www.imagingbook.com > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > Date: Wed, 24 Jul 2013 09:57:05 -0500 > From: Dinesh Joshi <[hidden email]> > Subject: Intensity analysis of a ROI from 2 channels > > Hell everyoneI want to measure the intensity of a region of interest (ROI) from 2 channels. > In the attached image I would like to measure the intensity of DAB in the regions of Neurofilaments (alexa fluor 488 fluorescence). Is there any way to pick a ROI in the merge image and get the intensity of that ROI from both DAB and neurofilament ? > > > Thanks > Dinesh C Joshi > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > Date: Wed, 24 Jul 2013 18:01:48 +0000 > From: "Bryan, Charles" <[hidden email]> > Subject: Measuring corrosion pit volumes. > > Hi All, > > I am new to ImageJ, and am trying to figure out how to measure the volumes of corrosion pits on polished metal surfaces. > > The raw data consists of image stacks collected with a laser confocal microscope, which are readily read and interpreted by ImageJ. The surface is defined by the layer with the brightest intensity. I need to be able to correct of surface tilt, determine the elevation in the stack for each x,y point, interpolate between adjacent layers to fill in holes (sometimes the pits walls are too steep, and leave shadows in the pit), and then calculate the pit volume and area. I think I can work out the individual steps if I have to, but suspect that the macros have already been developed, since this seems like a standard type of analysis. Does anyone have any suggestions? > > Thanks, > Charles Bryan > [hidden email] > > > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html > > ------------------------------ > > End of IMAGEJ Digest - 23 Jul 2013 to 24 Jul 2013 (#2013-216) > ************************************************************* > > -- > ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
«
Return to ImageJ
|
1 view|%1 views
| Free forum by Nabble | Edit this page |