Forums

Image Recognition based on defined images

Started by Prasun Dhara February 8, 2010
Hi All

I have a image database which has several known images with its label.Now I want to compare this images with a unknown input image.

If any known image is found in the unknown input image i need to label the image according to the image database.

For Example: I have 4 images A,B,C and D in my sample image database
Image label
A lion
B tiger
C cow
D man

Now i need to compare these images with a unknown image X and if any cow is found in that image i need to label the image as cow.

There are some more complex scenarios but i want to solve this first

Please can anyone help me regarding this ..I need the steps or algorithm to follow to solve this problem.

Thanks in advanced
Regards
Prasun Dhara
Prasun,
You can try using a simple cross correlation method, wherein, the unknown image is split up into frames equal to the size of the known template. You can then perform a cross correlation and look at the statistics to see if you have a match. The algorithm can be trained this way and a decision made based on the statistics.

Hope this helps.

Regards,
Nikhil.

Hi All
>
>I have a image database which has several known images with its label.Now I want to compare this images with a unknown input image.
>
>If any known image is found in the unknown input image i need to label the image according to the image database.
>
>For Example: I have 4 images A,B,C and D in my sample image database
>Image label
>A lion
>B tiger
>C cow
>D man
>
>Now i need to compare these images with a unknown image X and if any cow is found in that image i need to label the image as cow.
>
>There are some more complex scenarios but i want to solve this first
>
>Please can anyone help me regarding this ..I need the steps or algorithm to follow to solve this problem.
>
>Thanks in advanced
>Regards
>Prasun Dhara
>
>
That might help if your images are of different sizes. If they are the same size depicting the same object, why not simply compare each pixel value so that it outputs a 1 if the pixel at each position is of the same intensity. The sum of this comparison divided by the number of pixels in the image would be a 1 if all the pixels match perfectly. Otherwise it will be less than 1.
Again...provided all ur images are the same. Should they not be, then u could also probably look into analysing contours of the images, minimising the energy required to match them, or look into ASM algorithms to help you in your matching
Hope that answers ur questionsShail..

To: m...
From: n...@gmail.com
Date: Mon, 8 Feb 2010 16:57:31 -0500
Subject: [matlab] Re: Image Recognition based on defined images

Prasun,

You can try using a simple cross correlation method, wherein, the unknown image is split up into frames equal to the size of the known template. You can then perform a cross correlation and look at the statistics to see if you have a match. The algorithm can be trained this way and a decision made based on the statistics.

Hope this helps.

Regards,

Nikhil.

Hi All

>

>I have a image database which has several known images with its label.Now I want to compare this images with a unknown input image.

>

>If any known image is found in the unknown input image i need to label the image according to the image database.

>

>For Example: I have 4 images A,B,C and D in my sample image database

>Image label

>A lion

>B tiger

>C cow

>D man

>

>Now i need to compare these images with a unknown image X and if any cow is found in that image i need to label the image as cow.

>

>There are some more complex scenarios but i want to solve this first

>

>Please can anyone help me regarding this ..I need the steps or algorithm to follow to solve this problem.

>

>Thanks in advanced

>Regards

>Prasun Dhara