corner matching by SIFT [1 Attachment]

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

corner matching by SIFT [1 Attachment]

opencv-users mailing list
Hi all,

I am taking images taken from two different cameras.
The type of camera that i am using, is attached via this email.

When i got images. images taken from both camera give a lot of differences
in images,

Than i used corner matching to match harris corners of both images.
But corners does not match well because of difference in images.

I did research SIFT corner detector give good result in different images
taken from two camera and for corner matching, we are required some more
features such as descriptors. therefore i applied SIFT detector and
descriptors.

Now my question is how descriptors help to match corner points of both
images.
Corner points of both images are different because of difference in images.
than how corner matching can be accurate?
How descriptor helps to match the corners.

Can anyone elaborate it.

Thanks
Reply | Threaded
Open this post in threaded view
|

Re: corner matching by SIFT

opencv-users mailing list
Hi
  First of you have to estimate the planer homography using some standard
methodologies like Ransac, then go for corner detection




   Thanks & Regards
  Somnath Mukherjee
[hidden email]

On Tue, Apr 14, 2015 at 5:02 PM, Alrabi opencv [hidden email]
[OpenCV] <[hidden email]> wrote:

>
>  [Attachment(s) <#14cb7b3d45ab42fe_TopText> from Alrabi opencv included
> below]
>
>
> Hi all,
>
> I am taking images taken from two different cameras.
> The type of camera that i am using, is attached via this email.
>
> When i got images. images taken from both camera give a lot of differences
> in images,
>
> Than i used corner matching to match harris corners of both images.
> But corners does not match well because of difference in images.
>
> I did research SIFT corner detector give good result in different images
> taken from two camera and for corner matching, we are required some more
> features such as descriptors. therefore i applied SIFT detector and
> descriptors.
>
> Now my question is how descriptors help to match corner points of both
> images.
> Corner points of both images are different because of difference in
> images.
> than how corner matching can be accurate?
> How descriptor helps to match the corners.
>
> Can anyone elaborate it.
>
> Thanks
>
>  
>
Reply | Threaded
Open this post in threaded view
|

Re: corner matching by SIFT

opencv-users mailing list
Hello,

Descriptors calculated based on histogram local neighborhood of feature
points. You can take a euclidean distance between two descriptor vectors
for finding the matching measure. And mach those which gives less distance.
And reject feature point if you got more than one closed matched points. If
you got some matched points in a plane of object you can calculate
homography matrix between them and use ransac algorithm for outlier
rejection.


On Tue, Apr 14, 2015 at 6:20 PM, Somnath Mukherjee
[hidden email] [OpenCV] <[hidden email]> wrote:

>
>
> Hi
>   First of you have to estimate the planer homography using some standard
> methodologies like Ransac, then go for corner detection
>
>
>
>
>    Thanks & Regards
>   Somnath Mukherjee
> [hidden email]
>
> On Tue, Apr 14, 2015 at 5:02 PM, Alrabi opencv [hidden email]
> [OpenCV] <[hidden email]> wrote:
>
>>
>>  [Attachment(s) <#14cb7fb987ed98dd_14cb7b3d45ab42fe_TopText> from Alrabi
>> opencv included below]
>>
>>
>> Hi all,
>>
>> I am taking images taken from two different cameras.
>> The type of camera that i am using, is attached via this email.
>>
>> When i got images. images taken from both camera give a lot of
>> differences in images,
>>
>> Than i used corner matching to match harris corners of both images.
>> But corners does not match well because of difference in images.
>>
>> I did research SIFT corner detector give good result in different images
>> taken from two camera and for corner matching, we are required some more
>> features such as descriptors. therefore i applied SIFT detector and
>> descriptors.
>>
>> Now my question is how descriptors help to match corner points of both
>> images.
>> Corner points of both images are different because of difference in
>> images.
>> than how corner matching can be accurate?
>> How descriptor helps to match the corners.
>>
>> Can anyone elaborate it.
>>
>> Thanks
>>
>>
>  
>



--
Greetings

Shafeeq
Reply | Threaded
Open this post in threaded view
|

Re: corner matching by SIFT

opencv-users mailing list
so the homograph matrix will give me camera extrincsic and intrinsic
parameters .
right.. which i will use for image distortion and remapping and can compute
depth map.
Right?



On Tue, Apr 14, 2015 at 10:04 PM, shafeeq E [hidden email] [OpenCV] <
[hidden email]> wrote:

>
>
> Hello,
>
> Descriptors calculated based on histogram local neighborhood of feature
> points. You can take a euclidean distance between two descriptor vectors
> for finding the matching measure. And mach those which gives less distance.
> And reject feature point if you got more than one closed matched points. If
> you got some matched points in a plane of object you can calculate
> homography matrix between them and use ransac algorithm for outlier
> rejection.
>
>
> On Tue, Apr 14, 2015 at 6:20 PM, Somnath Mukherjee
> [hidden email] [OpenCV] <[hidden email]> wrote:
>
>>
>>
>> Hi
>>   First of you have to estimate the planer homography using some standard
>> methodologies like Ransac, then go for corner detection
>>
>>
>>
>>
>>    Thanks & Regards
>>   Somnath Mukherjee
>> [hidden email]
>>
>> On Tue, Apr 14, 2015 at 5:02 PM, Alrabi opencv [hidden email]
>> [OpenCV] <[hidden email]> wrote:
>>
>>>
>>>  [Attachment(s)
>>> <#14cb807bfb3ea6c0_14cb7fb987ed98dd_14cb7b3d45ab42fe_TopText> from
>>> Alrabi opencv included below]
>>>
>>>
>>> Hi all,
>>>
>>> I am taking images taken from two different cameras.
>>> The type of camera that i am using, is attached via this email.
>>>
>>> When i got images. images taken from both camera give a lot of
>>> differences in images,
>>>
>>> Than i used corner matching to match harris corners of both images.
>>> But corners does not match well because of difference in images.
>>>
>>> I did research SIFT corner detector give good result in different images
>>> taken from two camera and for corner matching, we are required some more
>>> features such as descriptors. therefore i applied SIFT detector and
>>> descriptors.
>>>
>>> Now my question is how descriptors help to match corner points of both
>>> images.
>>> Corner points of both images are different because of difference in
>>> images.
>>> than how corner matching can be accurate?
>>> How descriptor helps to match the corners.
>>>
>>> Can anyone elaborate it.
>>>
>>> Thanks
>>>
>>>
>>
>
>
> --
> Greetings
>
> Shafeeq
>
>  
>
Reply | Threaded
Open this post in threaded view
|

Re: corner matching by SIFT

opencv-users mailing list
Yes,

On Wed, Apr 15, 2015 at 7:12 AM, Alrabi opencv [hidden email]
[OpenCV] <[hidden email]> wrote:

>
>
> so the homograph matrix will give me camera extrincsic and intrinsic
> parameters .
> right.. which i will use for image distortion and remapping and can
> compute depth map.
> Right?
>
>
>
> On Tue, Apr 14, 2015 at 10:04 PM, shafeeq E [hidden email] [OpenCV] <
> [hidden email]> wrote:
>
>>
>>
>> Hello,
>>
>> Descriptors calculated based on histogram local neighborhood of feature
>> points. You can take a euclidean distance between two descriptor vectors
>> for finding the matching measure. And mach those which gives less distance.
>> And reject feature point if you got more than one closed matched points. If
>> you got some matched points in a plane of object you can calculate
>> homography matrix between them and use ransac algorithm for outlier
>> rejection.
>>
>>
>> On Tue, Apr 14, 2015 at 6:20 PM, Somnath Mukherjee
>> [hidden email] [OpenCV] <[hidden email]> wrote:
>>
>>>
>>>
>>> Hi
>>>   First of you have to estimate the planer homography using some
>>> standard methodologies like Ransac, then go for corner detection
>>>
>>>
>>>
>>>
>>>    Thanks & Regards
>>>   Somnath Mukherjee
>>> [hidden email]
>>>
>>> On Tue, Apr 14, 2015 at 5:02 PM, Alrabi opencv [hidden email]
>>> [OpenCV] <[hidden email]> wrote:
>>>
>>>>
>>>>  [Attachment(s)
>>>> <#14cbabe638530cac_14cb807bfb3ea6c0_14cb7fb987ed98dd_14cb7b3d45ab42fe_TopText>
>>>> from Alrabi opencv included below]
>>>>
>>>>
>>>> Hi all,
>>>>
>>>> I am taking images taken from two different cameras.
>>>> The type of camera that i am using, is attached via this email.
>>>>
>>>> When i got images. images taken from both camera give a lot of
>>>> differences in images,
>>>>
>>>> Than i used corner matching to match harris corners of both images.
>>>> But corners does not match well because of difference in images.
>>>>
>>>> I did research SIFT corner detector give good result in different
>>>> images taken from two camera and for corner matching, we are required some
>>>> more features such as descriptors. therefore i applied SIFT detector and
>>>> descriptors.
>>>>
>>>> Now my question is how descriptors help to match corner points of both
>>>> images.
>>>> Corner points of both images are different because of difference in
>>>> images.
>>>> than how corner matching can be accurate?
>>>> How descriptor helps to match the corners.
>>>>
>>>> Can anyone elaborate it.
>>>>
>>>> Thanks
>>>>
>>>>
>>>
>>
>>
>> --
>> Greetings
>>
>> Shafeeq
>>
>>
>  
>



--
Greetings

Shafeeq
Reply | Threaded
Open this post in threaded view
|

Re: corner matching by SIFT

opencv-users mailing list
Hi,

I am here with some more questions.

I read paper on Self Calibration of PTZ camera.
My question is that Is it possible to get the depth map by calibration only
single PTZ camera?

Another question is, As far as i understand, to perform the self
calibration on single camera or 2 camera, we need to get corner matching of
correspondens, which will give us the camera parameters i-e intrinsic and
extrinsic, but these parameters are of which camera?

My aim is to perform the self  calibration on two camera to make it stereo
camera and get the depth map.

Kindly guide me and share papers for self calibration of two cameras


On Wed, Apr 15, 2015 at 2:08 PM, shafeeq E [hidden email] [OpenCV] <
[hidden email]> wrote:

>
>
> Yes,
>
> On Wed, Apr 15, 2015 at 7:12 AM, Alrabi opencv [hidden email]
> [OpenCV] <[hidden email]> wrote:
>
>>
>>
>> so the homograph matrix will give me camera extrincsic and intrinsic
>> parameters .
>> right.. which i will use for image distortion and remapping and can
>> compute depth map.
>> Right?
>>
>>
>>
>> On Tue, Apr 14, 2015 at 10:04 PM, shafeeq E [hidden email] [OpenCV]
>> <[hidden email]> wrote:
>>
>>>
>>>
>>> Hello,
>>>
>>> Descriptors calculated based on histogram local neighborhood of feature
>>> points. You can take a euclidean distance between two descriptor vectors
>>> for finding the matching measure. And mach those which gives less distance.
>>> And reject feature point if you got more than one closed matched points. If
>>> you got some matched points in a plane of object you can calculate
>>> homography matrix between them and use ransac algorithm for outlier
>>> rejection.
>>>
>>>
>>> On Tue, Apr 14, 2015 at 6:20 PM, Somnath Mukherjee
>>> [hidden email] [OpenCV] <[hidden email]> wrote:
>>>
>>>>
>>>>
>>>> Hi
>>>>   First of you have to estimate the planer homography using some
>>>> standard methodologies like Ransac, then go for corner detection
>>>>
>>>>
>>>>
>>>>
>>>>    Thanks & Regards
>>>>   Somnath Mukherjee
>>>> [hidden email]
>>>>
>>>> On Tue, Apr 14, 2015 at 5:02 PM, Alrabi opencv [hidden email]
>>>> [OpenCV] <[hidden email]> wrote:
>>>>
>>>>>
>>>>>  [Attachment(s)
>>>>> <#14cd6985efd016bc_14cc1d75446e7ba7_14cbfd320cd69529_14cbb7a463e35c6c_14cbabe638530cac_14cb807bfb3ea6c0_14cb7fb987ed98dd_14cb7b3d45ab42fe_TopText>
>>>>> from Alrabi opencv included below]
>>>>>
>>>>>
>>>>> Hi all,
>>>>>
>>>>> I am taking images taken from two different cameras.
>>>>> The type of camera that i am using, is attached via this email.
>>>>>
>>>>> When i got images. images taken from both camera give a lot of
>>>>> differences in images,
>>>>>
>>>>> Than i used corner matching to match harris corners of both images.
>>>>> But corners does not match well because of difference in images.
>>>>>
>>>>> I did research SIFT corner detector give good result in different
>>>>> images taken from two camera and for corner matching, we are required some
>>>>> more features such as descriptors. therefore i applied SIFT detector and
>>>>> descriptors.
>>>>>
>>>>> Now my question is how descriptors help to match corner points of both
>>>>> images.
>>>>> Corner points of both images are different because of difference in
>>>>> images.
>>>>> than how corner matching can be accurate?
>>>>> How descriptor helps to match the corners.
>>>>>
>>>>> Can anyone elaborate it.
>>>>>
>>>>> Thanks
>>>>>
>>>>>
>>>>
>>>
>>>
>>> --
>>> Greetings
>>>
>>> Shafeeq
>>>
>>>
>>
>
>
> --
> Greetings
>
> Shafeeq
>
>  
>