imshow() pixelization in matplotlib 2.0.2 but not in 2.0.0

I'm plotting the same image (a numpy float32 array) with imshow(..., interpolation='bilinear') on matplotlib 2.0.0 and 2.0.2.

Click to see in full:

On 2.0.0 the image is perfectly smooth. However, on 2.0.2 there are noticeable pixelization effects.

Has imshow() changed between these two versions?

There have been some bugfixes, IIRC, mostly related to the handling of NaNs and masked points. Are there any in your data?

…

On Tue, Jun 6, 2017 at 10:54 AM, diogoff ***@***.***> wrote: I'm plotting the same image (a numpy float32 array) with imshow(..., interpolation='bilinear') on matplotlib 2.0.0 and 2.0.2. See the following figure: [image: matplotlib] <https://cloud.githubusercontent.com/assets/7391155/26835386/20980584-4ad0-11e7-8719-83beec5126d5.png> On 2.0.0 the image is perfectly smooth. However, on 2.0.2 there are noticeable pixelization effects. Has the imshow() routine changed? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8723>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-PvqiPeroQYvc-abNuKTIJk7LZINks5sBWhDgaJpZM4Nxdr0> .

Another possible source of issues *might* be colormap handling. It would be helpful to have a minimum working example to investigate with.

…

On Tue, Jun 6, 2017 at 11:23 AM, Benjamin Root ***@***.***> wrote: There have been some bugfixes, IIRC, mostly related to the handling of NaNs and masked points. Are there any in your data? On Tue, Jun 6, 2017 at 10:54 AM, diogoff ***@***.***> wrote: > I'm plotting the same image (a numpy float32 array) with imshow(..., > interpolation='bilinear') on matplotlib 2.0.0 and 2.0.2. > > See the following figure: > [image: matplotlib] > <https://cloud.githubusercontent.com/assets/7391155/26835386/20980584-4ad0-11e7-8719-83beec5126d5.png> > > On 2.0.0 the image is perfectly smooth. However, on 2.0.2 there are > noticeable pixelization effects. > > Has the imshow() routine changed? > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#8723>, or mute the > thread > <https://github.com/notifications/unsubscribe-auth/AARy-PvqiPeroQYvc-abNuKTIJk7LZINks5sBWhDgaJpZM4Nxdr0> > . >

Here is a MWE:

import matplotlib.pyplot as plt

data = [[ 0.001071,  0.032820,  0.105207,  0.127595,  0.147364,  0.124911,  0.073015,  0.000074, -0.003899, -0.001462],
        [ 0.000787,  0.031720,  0.107091,  0.183722,  0.202479,  0.204296,  0.154234,  0.085757,  0.026941, -0.003271],
        [ 0.001480, -0.000771,  0.016654,  0.214761,  0.278136,  0.304099,  0.246056,  0.169364,  0.085205, -0.001907],
        [ 0.000549, -0.000475,  0.001674,  0.247847,  0.345098,  0.378256,  0.332674,  0.229220,  0.133641,  0.053745],
        [ 0.000000,  0.002675,  0.001689,  0.266309,  0.414319,  0.459767,  0.425990,  0.317280,  0.190170,  0.099265],
        [ 0.000000,  0.001037, -0.002923,  0.254976,  0.408347,  0.529809,  0.532195,  0.406330,  0.272445,  0.163764],
        [ 0.000000,  0.000792, -0.013151,  0.216619,  0.397671,  0.567617,  0.616055,  0.508699,  0.379961,  0.263546],
        [ 0.000000,  0.002062, -0.016293,  0.157240,  0.368229,  0.613702,  0.712611,  0.639131,  0.505910,  0.338813],
        [ 0.000000,  0.001741,  0.003448,  0.100818,  0.333026,  0.611874,  0.787059,  0.773407,  0.644049,  0.450520],
        [ 0.000000, -0.001268, -0.008104,  0.061512,  0.294558,  0.604565,  0.814691,  0.865676,  0.776763,  0.557480],
        [ 0.000000, -0.001105, -0.002768,  0.036252,  0.252868,  0.544436,  0.775150,  0.920004,  0.890915,  0.690601],
        [ 0.000000, -0.001123, -0.003589,  0.028705,  0.194056,  0.471003,  0.728520,  0.921639,  0.969436,  0.773769],
        [ 0.000000, -0.000430, -0.004625,  0.017730,  0.149156,  0.423797,  0.677714,  0.912435,  0.984722,  0.846838],
        [ 0.000000,  0.000000, -0.002619,  0.036928,  0.136544,  0.410567,  0.666038,  0.905706,  0.942078,  0.894590],
        [ 0.000000, -0.000296, -0.003399,  0.050115,  0.189001,  0.451915,  0.696602,  0.869335,  0.893818,  0.826647],
        [ 0.000000,  0.000174, -0.003604,  0.083219,  0.248855,  0.517874,  0.739961,  0.843343,  0.803800,  0.718687],
        [-0.001720,  0.005825,  0.078039,  0.154466,  0.324385,  0.550980,  0.745479,  0.817428,  0.750877,  0.624806],
        [ 0.000719, -0.004020,  0.118178,  0.218264,  0.395383,  0.585784,  0.748400,  0.744401,  0.620376,  0.464464],
        [ 0.003257,  0.066374,  0.142013,  0.270861,  0.458675,  0.621705,  0.691592,  0.590108,  0.491103,  0.346859],
        [ 0.042411,  0.111464,  0.193880,  0.308119,  0.475352,  0.602842,  0.600717,  0.487950,  0.392807,  0.258806]]

plt.imshow(data, vmin=0., vmax=1., interpolation='bilinear')
plt.axis('off')
plt.show()

It is almost like it is treating those zeros as masked out, isn't it? attn: @tacaswell

…

On Tue, Jun 6, 2017 at 5:08 PM, diogoff ***@***.***> wrote: Here is a MWE: import matplotlib.pyplot as plt data = [[ 0.001071, 0.032820, 0.105207, 0.127595, 0.147364, 0.124911, 0.073015, 0.000074, -0.003899, -0.001462], [ 0.000787, 0.031720, 0.107091, 0.183722, 0.202479, 0.204296, 0.154234, 0.085757, 0.026941, -0.003271], [ 0.001480, -0.000771, 0.016654, 0.214761, 0.278136, 0.304099, 0.246056, 0.169364, 0.085205, -0.001907], [ 0.000549, -0.000475, 0.001674, 0.247847, 0.345098, 0.378256, 0.332674, 0.229220, 0.133641, 0.053745], [ 0.000000, 0.002675, 0.001689, 0.266309, 0.414319, 0.459767, 0.425990, 0.317280, 0.190170, 0.099265], [ 0.000000, 0.001037, -0.002923, 0.254976, 0.408347, 0.529809, 0.532195, 0.406330, 0.272445, 0.163764], [ 0.000000, 0.000792, -0.013151, 0.216619, 0.397671, 0.567617, 0.616055, 0.508699, 0.379961, 0.263546], [ 0.000000, 0.002062, -0.016293, 0.157240, 0.368229, 0.613702, 0.712611, 0.639131, 0.505910, 0.338813], [ 0.000000, 0.001741, 0.003448, 0.100818, 0.333026, 0.611874, 0.787059, 0.773407, 0.644049, 0.450520], [ 0.000000, -0.001268, -0.008104, 0.061512, 0.294558, 0.604565, 0.814691, 0.865676, 0.776763, 0.557480], [ 0.000000, -0.001105, -0.002768, 0.036252, 0.252868, 0.544436, 0.775150, 0.920004, 0.890915, 0.690601], [ 0.000000, -0.001123, -0.003589, 0.028705, 0.194056, 0.471003, 0.728520, 0.921639, 0.969436, 0.773769], [ 0.000000, -0.000430, -0.004625, 0.017730, 0.149156, 0.423797, 0.677714, 0.912435, 0.984722, 0.846838], [ 0.000000, 0.000000, -0.002619, 0.036928, 0.136544, 0.410567, 0.666038, 0.905706, 0.942078, 0.894590], [ 0.000000, -0.000296, -0.003399, 0.050115, 0.189001, 0.451915, 0.696602, 0.869335, 0.893818, 0.826647], [ 0.000000, 0.000174, -0.003604, 0.083219, 0.248855, 0.517874, 0.739961, 0.843343, 0.803800, 0.718687], [-0.001720, 0.005825, 0.078039, 0.154466, 0.324385, 0.550980, 0.745479, 0.817428, 0.750877, 0.624806], [ 0.000719, -0.004020, 0.118178, 0.218264, 0.395383, 0.585784, 0.748400, 0.744401, 0.620376, 0.464464], [ 0.003257, 0.066374, 0.142013, 0.270861, 0.458675, 0.621705, 0.691592, 0.590108, 0.491103, 0.346859], [ 0.042411, 0.111464, 0.193880, 0.308119, 0.475352, 0.602842, 0.600717, 0.487950, 0.392807, 0.258806]] plt.imshow(data, vmin=0., vmax=1., interpolation='bilinear') plt.axis('off') plt.show() [image: matplotlib] <https://user-images.githubusercontent.com/7391155/26851969-802a9fb4-4b04-11e7-8d29-4d49e3df3c30.png> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#8723 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-Btlf7KfBsDWGii3gi5L-MELxm0Yks5sBb-zgaJpZM4Nxdr0> .

The pixelization appears around the negative values which are being clipped by vmin=0.

you are right. It is the pixels that fall below vmin. There was some recent work around vmin/vmax. maybe that had some unintended consequences?

…

On Tue, Jun 6, 2017 at 5:14 PM, diogoff ***@***.***> wrote: The pixelization appears around the negative values which are being clipped by vmin=0. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#8723 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-D53Gn1CAvCdzYvawoSBkw6yQ4YQks5sBcElgaJpZM4Nxdr0> .

It seems that when those (negative) pixels are being clipped, the interpolation no longer works as it used to.
Thanks for your help.

Possible culprit: #8024

…

On Tue, Jun 6, 2017 at 5:22 PM, diogoff ***@***.***> wrote: It seems that when those (negative) pixels are being clipped, the interpolation no longer works as it used to. Thanks for your help. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#8723 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-NkZqs3thlZAIm4n-tE0ITMSxu_Qks5sBcL-gaJpZM4Nxdr0> .

Yep, this bisects to 1a7c4ef

This is a duplicate of #8631 which has a proposed solution in the comments.

Closed by #8966

Navigation Menu

Repository navigation

imshow() pixelization in matplotlib 2.0.2 but not in 2.0.0 #8723

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

imshow() pixelization in matplotlib 2.0.2 but not in 2.0.0 #8723

Description

Activity

WeatherGod commented on Jun 6, 2017

WeatherGod commented on Jun 6, 2017

diogoff commented on Jun 6, 2017

WeatherGod commented on Jun 6, 2017

diogoff commented on Jun 6, 2017

WeatherGod commented on Jun 6, 2017

diogoff commented on Jun 6, 2017

WeatherGod commented on Jun 6, 2017

dstansby commented on Jun 15, 2017

tacaswell commented on Jun 17, 2017

tacaswell commented on Aug 31, 2017

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions