how can I start coding CBIR using color histogram

Hi all

I need help in how could I make a simple CBIR system using histogram technique

from where i should start

[Nivedita Nagappa]

Hi...I ve a similar problem..and im totally new to Matlab...can anyone pls help me out in this..??

PS: I too want to develop a simple CBIR system using color histogram

[Image Analyst]

@Ragini Gaikwad, this is not a comment for @shima said on regarding his 8 year old question on CBIR. This is a new question, so please start your own thread.

[Ragini Gaikwad]

sorry for that..it was a mistake

Answer 1

[Image Analyst]

You mean like this: http://labs.tineye.com/multicolr/ ?

Just try taking the mean of each color channel and have that be your feature vector that you save along with each image's record in your data base.

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);
% Get means.
meanRed = mean2(redChannel);
meanGreen = mean2(greenChannel);
meanBlue = mean2(blueChannel);

Then, for a desired color that you want to find in your database, search all records in the database for those whose mean colors are close enough to your desired color and return them.

[shima said]

hi again I used grayscale and i stuck when i compute Euclidean distance after i get it so what should i do how to retreive the similarity image in my database one qurey image and file image which have 111 images

[Image Analyst]

Just sort the distances and take the ones with the smallest difference first.

[Walter Roberson]

Find the indices of the images with the lowest Euclidean distance. Retrieve them however needed to retrieve an arbitrary image.

In your current code, how do you refer to (say) the 9'th image? The 37'th ?

[shima said]

sorry , I didn't get what u meant both of you after I stort Euclidean distance or find the lowest what should I do i stuck there couldanyone to hlep me to write that in matlab

[Image Analyst]

The square root of the sum of the squares of the distances:

euclideanDistances = sqrt(sum((featureVector1 - featureVector2))^2));

or something like that.