[IPOL discuss] BM3D ready in dev. server

[Jean-Michel Morel]

Dear all,

We need to minimize the maintenance risk and effort and our demo system. 
I try to contribute to it by enforcing standard demo settings.

As a rule of thumb, when you design a new demo, the first question is: 
is the demo type similar or identical to an already existing standard 
demo?

In this case which is frequent, I propose to leave no other choice to 
the redactor than a copy-paste of the former ones with the minimal 
adaptation.

Or, if the redactor thinks that a new setting is widely preferable, then 
this change should be accepted by all and applied by the redactor to all 
equivalent demos, not just the one he is in charge of.

This exigence comes up naturally for scientific reasons. All equivalent 
algorithms must be tested by users in exactly the same way.

We are preparing groups of demos,  "benchmarks", which must therefore 
have EXACTLY the same settings. Their demo sites should look identical 
to the users.

I enclose the current list of projected benchmarks below. Each one is 
expanding rapidly. I invite you eagerly to fix a unique format for each 
benchmark. If several editors take care of one benchmark, they should 
arrive at a final proposition.


We are currently preparing seven such benchmarks. I enclose their list 
of demos and the current redactors

A) Denoising : José Luis, Miguel, Pascal Getreuer
ys_dct_denoising
bcm_non_local_means_denoising
g_tv_denoising
+ BM3D
+ NL-means PCA
+ KSVD

B) Demosaicking: Pascal Getreuer

bcms_self_similarity_driven_demosaicking
g_demosaicking_with_contour_stencils
g_lian_adaptive_filtering_demosaicking
g_malvar_he_cutler_linear_image_demosaicking
g_zhang_wu_lmmse_image_demosaicking

C) Example based texture synthesis Nicolas, ???
ggm_random_phase_texture_synthesis
+ Heeger Bergen
+ Wei Levoy
+ Portilla Simoncelli

D) Image interpolation  Pascal Getreuer
g_image_interpolation_with_contour_stencils
g_interpolation_geometric_contour_stencils
g_linear_methods_for_image_interpolation
g_roussos_diffusion_interpolation

E) Epipolar rectification Pascal Monasse, Gabriele
m_quasi_euclidean_epipolar_rectification
+ 3 more

F) Various color-contrast adjustment  José Luis
g_localcolorcorrection
lmps_selective_contrast_adjustment
color balance
+ histogram piecewise linear
+ gamma correction

G) Stéréo  Pascal Monasse, Gabriele
ltp_stereovision_with_graph_cuts
a_contrario_block_matching
+ several block matching
+ several masks

H) Segmentation

I) gestalt detection
gjmr_line_segment_detector





rafael grompone von gioi a écrit :
> Hello!
> 
> I have a question:
> If I understand the demo, it adds noise to the input image,
> and then tries to remove it. Why can't we also ask
> the demo to remove just the noise that is already on the image?
> 
> And an error:
> I tried an image that is 501x720 pixels.
> It seems that the result page is designed for a fixed size of
> image: The "Measure" information and the "This result is archived..."
> text appears behind the image, that seems to big in the Y axis.
> 
> Then a suggestion:
> I think it would be nice to have at result page
> the extracted noise shown as a fourth image.
> 
> rafa
> 
> 
> 
> On Thu, Apr 28, 2011 at 12:09 PM, Miguel Colom
> <Miguel.Colom at cmla.ens-cachan.fr> wrote:
>> Hello everybody,
>> the BM3D demo is now ready in the dev. server. You can use it and report
>> errors if you find any.
>>
>> The URL of the demo is:
>> http://dev.ipol.im/~colom/ipol_demo/bm3d/
>>
>> Best,
>> Miguel
>>
>> _______________________________________________
>> discuss mailing list
>> discuss at list.ipol.im
>> http://tools.ipol.im/mailman/listinfo/discuss
>>
>>
> _______________________________________________
> discuss mailing list
> discuss at list.ipol.im
> http://tools.ipol.im/mailman/listinfo/discuss
>