texture detection

[Madhubalan]

Hi

Local binary pattern desciption:

During the precise target object detection step, LBP feature distributions are generally used. A LBP code can be calculated using neighborhood intensities. Generally, if the neighborhood pixel value is greater than the center pixel value, a 1 is assigned to the code, otherwise a 0 is assigned. 3x3 neighborhood pixels were used, as shown in Fig. 3. The neighborhood pixels wereassigned by comparing them with the value of the center pixel:

Then, 8-bits were used to obtain the LBP value of the
center pixel:

Every pixel was encoded into an 8-bit LBP value. The histogram of the LBP image was calculated by counting the number of pixels with a certain LBP value.

How to implement this in labview?

Thank you

[Klemen]

Hello,

for the first part take a sliding window (kernel) of 3x3 size (9 elements) and move it through your image. Before that, reserve an image in memory of the same size and set the pixels to the calculated values from your original image, based on the first equation (intensities of surrounding pixels with respect to the central pixel).

Best regards,

K

[Madhubalan]

Hi,

     can please provide me with some example code for local binary pattern 

Thank you

[Klemen]

Hello,

here is a very good example code for LBP and Extended LBP with explanation:

http://www.bytefish.de/blog/local_binary_patterns/

You can also search for more examples (i found some in Matlab) and maybe read the original paper of the authors about LBP.

This is also very interesting to me, but i really do not have the time to go further into this now. But if you are able to create a lv code, you are welcome to post it here - it would be very usefull to me. Thanks.

Best regards,

K

[vekkuli]

I'm not entirely familiar with the LBT but based on the information here I made this vi. If it seems like the kind of thing you are looking for we can work on it further to make it suitable for your application. I hope it's any help!

[Madhubalan]

Hi Vekkuli,

                   I am using the technique as per this pdf.If you can look at this pdf,you can help me.

Thank you

[vekkuli]

I see. Well, in case I didn't make any silly errors the VI should be on the right track with the LBP part. It creates the 3x3 neighbourhood LBP 2D Array from the 8-bit grayscale source image.

[Madhubalan]

Hi Vekkuli, 

                   It is working fine.Thank you for the LBP part,it would be great if you help in LBP histogram matching.

Thank you

[vekkuli]

It really depends how you setup the matching/comparison scenario. The measurement itself is really simple, I attached a VI of the histogram intersection calculation as described in the document.

[Madhubalan]

Hi vekkuli,

             Suppose if i want to compare two similar image(attached file),and in the ouput i have to display the image by highlighting the identical objects by comparing the two iput images using histogram intersection

in the pdf you see the image file under fig5.

(d) The results of color histogram matching (rectangular boxes represent object target candidates),

Thank you