Mailman 3 April 2017 - scikit-image

Re: [scikit-image] Fwd: SSIM with tolerances
by mat saunders 04 Apr '17

04 Apr '17

Hi Stefan, Thank you for feedback I will look at it more closely and see what I can come up with. I did not know about the matplotlib comparison so I will see what they do. I did come across this the other day and wondered about down scaling. http://tech.jetsetter.com/2017/03/21/duplicate-image-detection/ Thanks, Mat Saunders On Mon, Apr 3, 2017 at 3:47 PM, Stefan van der Walt <stefanv(a)berkeley.edu> wrote: > Hi Mat > > You'll probably have to come up with some heuristics for the kinds of > errors you wish to identify. As a very crude first attempt, you may try to > downscale the images before comparison to reduce sensitivity. Perhaps also > take a look at what Matplotlib does in their test suite to compare > different renderings of plots (I know they do image comparison too). > > Best regards > Stéfan > > > On Mon, Apr 3, 2017, at 12:00, mat saunders wrote: > > I sent this to scikit-learn by accident, so forwarding to image. > > Hi, > > I am using SSIM to compare 2 video streams\sets of images and I find it to > be almost too accurate. I would like some fudge factor like other image > comparison tools have. I used to do it in an automated test suite but due > to file sizes and amounts I turned to scikit. > > I do quality assurance on a render engine and we just want to make sure > the images are meaningfully identical build to build. Currently with SSIM I > am seeing things as small as 4 pixels across a 1920x1080 image different. I > personally would like to ignore those 4 pixels but still catch meaningful > items. Say if 8 pixels near each other were off keep those but if they are > 8 pixels randomly through the image ignore them. > > Does this sound like something logical, say using an adjacency of pixels > with a tolerance value for color and number of pixels as arguments? > > See attached image for example of how little is different in the entire > image. It is a GIF zoomed in to the exact spot of 3 different pixels so > hopefully it works. I also attached the comparison image to show how little > changed. > > Another issue that we had is we had to turn off AA to get rid of noise, it > is amazing how accurate this library is!! > > Regards, > Mathew Saunders > *_______________________________________________* > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image > > Email had 2 attachments: > > - 3dots-again.gif > 138k (image/gif) > - Pasted image at 2017_03_28 12_51 PM.png > 560k (image/png) > > > > _______________________________________________ > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image > >

1 0

Re: [scikit-image] Fwd: SSIM with tolerances
by Stefan van der Walt 04 Apr '17

04 Apr '17

Hi Mat You'll probably have to come up with some heuristics for the kinds of errors you wish to identify. As a very crude first attempt, you may try to downscale the images before comparison to reduce sensitivity. Perhaps also take a look at what Matplotlib does in their test suite to compare different renderings of plots (I know they do image comparison too). Best regards Stéfan On Mon, Apr 3, 2017, at 12:00, mat saunders wrote: > I sent this to scikit-learn by accident, so forwarding to image. > > Hi, > > I am using SSIM to compare 2 video streams\sets of images and I find > it to be almost too accurate. I would like some fudge factor like > other image comparison tools have. I used to do it in an automated > test suite but due to file sizes and amounts I turned to scikit. > > I do quality assurance on a render engine and we just want to make > sure the images are meaningfully identical build to build. Currently > with SSIM I am seeing things as small as 4 pixels across a 1920x1080 > image different. I personally would like to ignore those 4 pixels but > still catch meaningful items. Say if 8 pixels near each other were > off keep those but if they are 8 pixels randomly through the image > ignore them. > > Does this sound like something logical, say using an adjacency of > pixels with a tolerance value for color and number of pixels as > arguments? > > See attached image for example of how little is different in the > entire image. It is a GIF zoomed in to the exact spot of 3 different > pixels so hopefully it works. I also attached the comparison image to > show how little changed. > > Another issue that we had is we had to turn off AA to get rid of > noise, it is amazing how accurate this library is!! > > Regards, > Mathew Saunders > _________________________________________________ > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image > Email had 2 attachments: > * 3dots-again.gif 138k (image/gif) > * Pasted image at 2017_03_28 12_51 PM.png 560k (image/png)

1 0

Fwd: SSIM with tolerances
by mat saunders 04 Apr '17

04 Apr '17

I sent this to scikit-learn by accident, so forwarding to image. Hi, I am using SSIM to compare 2 video streams\sets of images and I find it to be almost too accurate. I would like some fudge factor like other image comparison tools have. I used to do it in an automated test suite but due to file sizes and amounts I turned to scikit. I do quality assurance on a render engine and we just want to make sure the images are meaningfully identical build to build. Currently with SSIM I am seeing things as small as 4 pixels across a 1920x1080 image different. I personally would like to ignore those 4 pixels but still catch meaningful items. Say if 8 pixels near each other were off keep those but if they are 8 pixels randomly through the image ignore them. Does this sound like something logical, say using an adjacency of pixels with a tolerance value for color and number of pixels as arguments? See attached image for example of how little is different in the entire image. It is a GIF zoomed in to the exact spot of 3 different pixels so hopefully it works. I also attached the comparison image to show how little changed. Another issue that we had is we had to turn off AA to get rid of noise, it is amazing how accurate this library is!! Regards, Mathew Saunders

1 0

How to see full image after transformation not the cropped one
by Serge Shakhov 03 Apr '17

03 Apr '17

Hello everybody This question was raised here: Google Grup | | | Google Grup Google Grup memungkinkan Anda untuk membuat dan berpartisipasi dalam forum online serta grup berbasis email deng... | | | × × × But I didn't find an answer.I'm doing 2D piecewise affine transformation with skimage.transform.PiecewiseAffineTransformResult image is heavily cropped.I tried to play with output_shape, mode and clip parameters but without any effect, image is still cropped.Could anyone point me what am I doing wrong?Thanks Best regards, Serge Shakhov.

3 2