Re: [libdmtx-open_discussion] reed-solomon decoding

[François A. <fra...@gm...>]

Hi Mike,

Thanks alot for your answer. Unfortunately I could not reproduce the exact
error with libdmtx. However, I realized that this problem comes from the
fact that I removed  the assert line in the following without replacing by a
proper catch...(mea culpa :) )

PushOutputC40TextWord(DmtxMessage *msg, C40TextState *state, int value)
{
   *assert(value >= 0 && value < 256);*

      ...
}

Indeed, because I wrote the code localization myself and then "feeds" the
corner position to the libdmtx engine, due to some sampling errors coming
from this localization, I end up with some false CWs (0x00...for instance).
In that case, the RS decoder ends up with a decodePass but a totally exotic
string!

So far, catching these wrong values prevent me from outputting again the
false codes.

Also, changing the max allowed error number seemed to fix the erroneous
behavior.

Thanks again for your help


François



On Mon, Oct 26, 2009 at 10:53 PM, Mike Laughton <mi...@dr...>wrote:

> Hi François,
>
> Thank you for your email. Please see my comments below:
>
> Is it possible to get by chance a decodable code from a random sampling? In
>> my opinion, the probabilities are really small for that isn't it?
>
>
> Yes, the odds of this should be quite small but not completely impossible.
>
> I recently observed that sometimes, when trying to decode an image with
>> wrong parameters (for instance setting the size to be 12x26 instead of
>> 16x16), I receive a decodePass result from the rs_decode_char which in turn
>> leads to an exotic result string.
>>
>
> Can you provide a sample image along with the conditions that cause this to
> occur? I would be very interested in reproducing this condition for further
> troubleshooting.
>
> In the meantime if you are concerned about this issue you can try adjusting
> the number of allowable errors, since this problem will likely go away if
> you only allow a small number of corrections.
>
> Mike
>
>
> --
> Mike Laughton
> Chief Software Architect
> Dragonfly Logic, Inc.
> 612.845.3339 (m)
>



-- 
---------------------------------------------------------------
François Aspert
9 rue Marterey
1005 Lausanne
Switzerland
Phone: +41 21 311 70 27
Email: fra...@gm...
---------------------------------------------------------------
Sent from Zurich, ZH, Switzerland