Hi All,

I have an image which is very blocky (image can be seen athttp://img222.imageshack.us/my.php?image=degradedimagesh2.png) and
would like to try and improve it's quality as much as possible.

I understand that the image is of such a poor quality that I do not
expect any huge improvements - that is not my aim, my aim is to see if
it can be improved at all, and if so, which method(s) could yield the
best results.

So far I have tried some iterative restoration techniques such as
Richardson-Lucy deconvolution and Blind deconvolution - both have
proved to give a slight improvement.

I was hoping someone would have some other suggestions as to how I
could possibly improve the image.

Just in case understanding how the image was degraded in the first
place might help lead to a good suggestion for improvement algorithms,
I created the image using the following steps:

* 8x8 Sub-Block and DCT image
* store top 2 coefficients from each sub-block
* use these to create a new image

illywhacker wrote:
This is certainly a homework problem, and I was disposed to help you,
but then all those pop-ups came up...

Perhaps the OP should use photobucket.

Very nice hosting IMHO.

Arbitrary link sample:

http://i48.photobucket.com/albums/f234/bugbear33/wdibb.jpg

BugBear

Hi All,

I have an image which is very blocky (image can be seen athttp://img222.imageshack.us/my.php?image=degradedimagesh2.png) and
would like to try and improve it's quality as much as possible.

I understand that the image is of such a poor quality that I do not
expect any huge improvements - that is not my aim, my aim is to see if
it can be improved at all, and if so, which method(s) could yield the
best results.

So far I have tried some iterative restoration techniques such as
Richardson-Lucy deconvolution and Blind deconvolution - both have
proved to give a slight improvement.

I was hoping someone would have some other suggestions as to how I
could possibly improve the image.

Just in case understanding how the image was degraded in the first
place might help lead to a good suggestion for improvement algorithms,
I created the image using the following steps:

* 8x8 Sub-Block and DCT image
* store top 2 coefficients from each sub-block
* use these to create a new image

Many thanks,

AJ

AJ wrote:
Hi All,

I have an image which is very blocky (image can be seen at
http://img222.imageshack.us/my.php?image=degradedimagesh2.png) and
would like to try and improve it's quality as much as possible.

I understand that the image is of such a poor quality that I do not
expect any huge improvements - that is not my aim, my aim is to see if
it can be improved at all, and if so, which method(s) could yield the
best results.

So far I have tried some iterative restoration techniques such as
Richardson-Lucy deconvolution and Blind deconvolution - both have
proved to give a slight improvement.

I was hoping someone would have some other suggestions as to how I
could possibly improve the image.

You could play the match boundary continuity game with a cosmetic
enhancement. I think the IJG codec can be persuaded to do this with
suitable options.



Just in case understanding how the image was degraded in the first
place might help lead to a good suggestion for improvement algorithms,
I created the image using the following steps:

* 8x8 Sub-Block and DCT image
* store top 2 coefficients from each sub-block
* use these to create a new image

So your reconstruction is from DC and lowest frequency horizontal
component only? I don't think you can do much starting from there.

Taking a subset square of low frequency coefficients around the DC
component would give you a classic undersampling problem. eg 2x2 or 3x3
measured components. But unless the reconstructed image clips against 0
or 255 you don't get much by way of clues to regain resolution lost by
the downsampling. And your test image is well clear of either.

A simple field of white dots on a pure black ground would give you a
test image where you might stand a better chance of success.

In an ideal world maxent or another optimised nonlinear reconstruction
might hope to extrapolate resolution to 3x the highest measured
frequency for regions with the best signal to noise.

Regards,
Martin Brown
** Posted fromhttp://www.teranews.com**

On Jun 9, 5:38 am, AJ <arandal... at (no spam) gmail.com> wrote:



Hi All,

I have an image which is very blocky (image can be seen athttp://img222.imageshack.us/my.php?image=degradedimagesh2.png) and
would like to try and improve it's quality as much as possible.

I understand that the image is of such a poor quality that I do not
expect any huge improvements - that is not my aim, my aim is to see if
it can be improved at all, and if so, which method(s) could yield the
best results.

So far I have tried some iterative restoration techniques such as
Richardson-Lucy deconvolution and Blind deconvolution - both have
proved to give a slight improvement.

I was hoping someone would have some other suggestions as to how I
could possibly improve the image.

Just in case understanding how the image was degraded in the first
place might help lead to a good suggestion for improvement algorithms,
I created the image using the following steps:

* 8x8 Sub-Block and DCT image
* store top 2 coefficients from each sub-block
* use these to create a new image

Many thanks,

AJ

Is this an improvement?:

http://www.general-cathexis.com/images/degradedimagesh2_2.png

I modeled your image as an 8X nearest neighbor (box) enlargement in
both horizontal and vertical directions. All variation within a block
is horizontal. In other words, the box model is exact only in the
vertical direction. I reduced 0.125 X using box kernel and then
enlarged 8X using a proprietary nonlinear enlargement method available
in SAR Image Processor.

This is certainly a homework problem, and I was disposed to help you,
but then all those pop-ups came up...

I'm impressed!  The edges are very crisp, much better than I am
currently getting.  Thank you for going to the effort of experimenting
with it and uploading your results.

It would be great if I could use my results (iterative restoration and
interpolation) and then apply your technique to edges, as I think that
the "inner" details such as the face look quite good with my
restoration technique, and the edges of the image look much better
with yours.

I would be really grateful if you could expand on the process you took
to achieve this image, such as the "8X nearest neighbor (box)
enlargement".

Did you implement this in Matlab?  I'm not sure what
you mean by this...is it a filter you have applied?

Hi All,

I have an image which is very blocky (image can be seen at
http://img222.imageshack.us/my.php?image=degradedimagesh2.png ) and
would like to try and improve it's quality as much as possible.

I understand that the image is of such a poor quality that I do not
expect any huge improvements - that is not my aim, my aim is to see if
it can be improved at all, and if so, which method(s) could yield the
best results.

So far I have tried some iterative restoration techniques such as
Richardson-Lucy deconvolution and Blind deconvolution - both have
proved to give a slight improvement.

I was hoping someone would have some other suggestions as to how I
could possibly improve the image.

Just in case understanding how the image was degraded in the first
place might help lead to a good suggestion for improvement algorithms,
I created the image using the following steps:

* 8x8 Sub-Block and DCT image
* store top 2 coefficients from each sub-block
* use these to create a new image

Interspersed comments:

On Jun 9, 11:57 am, AJ <arandal... at (no spam) gmail.com> wrote:



I'm impressed! The edges are very crisp, much better than I am
currently getting. Thank you for going to the effort of experimenting
with it and uploading your results.

It would be great if I could use my results (iterative restoration and
interpolation) and then apply your technique to edges, as I think that
the "inner" details such as the face look quite good with my
restoration technique, and the edges of the image look much better
with yours.

The edge preserving method is not post processing and is proprietary.



I would be really grateful if you could expand on the process you took
to achieve this image, such as the "8X nearest neighbor (box)
enlargement".

http://en.wikipedia.org/wiki/Nearest_neighbor_interpolation

I didn't produce any box enlargement. I pretended that your image was
a box enlargement of a hypothetical image 1/8 its size. Thusly, only
for my convenience, I threw away all information contained in your DCT
coefficients above 0 frequency. With respect to reduction, an
interpolation kernel is typically expanded by a factor inversely
proportional to the scale. More specifically, the pixels in each of
your 8x8 blocks were simply averaged to produce one pixel in a reduced
image. Then that reduced image was enlarged by my proprietary method
which incorporates prior knowledge that the image was a box reduction.

Did you implement this in Matlab? I'm not sure what
you mean by this...is it a filter you have applied?

No, I implemented SAR Image Processor in C++. You can download a free
save disabled demo to play with. Google it.

On Jun 9, 10:30 pm, aruzinsky <aruzin... at (no spam) general-cathexis.com> wrote:





Interspersed comments:

On Jun 9, 11:57 am, AJ <arandal... at (no spam) gmail.com> wrote:

I'm impressed!  The edges are very crisp, much better than I am
currently getting.  Thank you for going to the effort of experimenting
with it and uploading your results.

It would be great if I could use my results (iterative restoration and
interpolation) and then apply your technique to edges, as I think that
the "inner" details such as the face look quite good with my
restoration technique, and the edges of the image look much better
with yours.

The edge preserving method is not post processing and is proprietary.

I would be really grateful if you could expand on the process you took
to achieve this image, such as the "8X nearest neighbor (box)
enlargement".

http://en.wikipedia.org/wiki/Nearest_neighbor_interpolation

I didn't produce any box enlargement.  I pretended that your image was
a box enlargement of a hypothetical image 1/8 its size.   Thusly, only
for my convenience, I threw away all information contained in your DCT
coefficients above 0 frequency.  With respect to reduction, an
interpolation kernel is typically expanded by a factor inversely
proportional to the scale.  More specifically, the pixels in each of
your 8x8 blocks were simply averaged to produce one pixel in a reduced
image.  Then that reduced image was enlarged by my proprietary method
which incorporates prior knowledge that the image was a box reduction.

Did you implement this in Matlab?  I'm not sure what
you mean by this...is it a filter you have applied?

No, I implemented SAR Image Processor in C++.  You can download a free
save disabled demo to play with.  Google it.

Thanks for your reply.

I did find SAR Image Processor after I had responded to your last
message.  However, I am running a Linux OS so unfortunately cannot run
it.

I did however use interpolation to improve the quality of my image, so
thank you for that suggestion.  I have uploaded my "best" attempt at
restoring the image, and it can be seen at:

http://i287.photobucket.com/albums/ll146/arandall85/interpolated_iter...

To get to this point, I have used blind deconvolution, and then
interpolated the result.  I'm quite happy with this outcome, although
it is a little blurry.  What do you all think?

Ideally, I would like to "fix" the edges in some way, as they are
still a bit blocky, and I would like to sharpen the whole image
slightly - but I think I may be hoping for a bit too much! ;)

If anyone has any further suggestions to improve this image, I would
be really grateful.  Nearly there...

Kind regards,

AJ- Hide quoted text -

- Show quoted text -

On Jun 9, 5:43 pm, AJ <arandal... at (no spam) gmail.com> wrote:

http://i287.photobucket.com/albums/ll146/arandall85/interpolated_iter...

To get to this point, I have used blind deconvolution, and then
interpolated the result. I'm quite happy with this outcome, although
it is a little blurry. What do you all think?

Ideally, I would like to "fix" the edges in some way, as they are
still a bit blocky, and I would like to sharpen the whole image
slightly - but I think I may be hoping for a bit too much! ;)

If anyone has any further suggestions to improve this image, I would
be really grateful. Nearly there...

Kind regards,

AJ- Hide quoted text -

- Show quoted text -

But, although similar, this is not a deconvolution problem. Like
convolution, you have a linear transformation which can be put in
matrix form,

Ax = y

and you want to estimate x from y, but, unlike convolution, most of
the eigenvalues of A are zero and (I am uncertain so check me) A is
idempotent, i.e., A^2 = A. In low rank cases, it is generally good
practice to constrain a solution, x', so that Ax' = y. I checked your
result and it violates this constraint because the averages of the 8x8
blocks in your result are not equal to those of your input image.

...
http://www.general-cathexis.com/images/degradedimagesh2_3.png

, which is one of an infinite number of possible solutions, x', for
Ax' = y.- Hide quoted text -

- Show quoted text -

and you want to estimate x from y, but, unlike convolution, most of
the eigenvalues of A are zero and (I am uncertain so check me) A is
idempotent, i.e., A^2 = A.   In low rank cases, it is generally good
practice to constrain a solution, x', so that Ax' = y.  I checked your
result and it violates this constraint because the averages of the 8x8
blocks in your result are not equal to those of your input image.

You shouldn't expect them to be exactly equal.

That is overconstrained.

The best you can hope for is that the DC and First AC component (which
is all he has) computed from the model image are on average within half
the quantisation error of the JPEG encoding in a least squares test.