How to get fixed dimension feature vectors for different size boxes?

Hi, In my detection task I have bounding box for each training image and those bounding boxes are different sizes for different images. How to get fixed dimension feature vectors for different size boxes? I am using Contourlet transform.

Answer 1

[Image Analyst]

Maybe they should be different sizes. If you're tracking/detecting fish in a fish tank, then I can see that the bounding boxes will be different sizes from one image to the next, even in the same fish tank.

To get actual real world lengths and areas, you need to image an object of known dimension, such as a ruler. See attached demo.

[Image Analyst]

And it depends on what is in your feature vector. If it's just mean intensity and area, then yeah, you could have just 2 measurements regardless of the picture size, but if the feature vector is has the centroids of numerous blobs in the box, then it could change on a box-by-box basis because different boxes could have different number of centroids in them. (Example being a box around a word but the word can have different numbers of letters.)

[FARHAD]

Thanks for your answer. Actually I am detecting a car and using Contourlet decomposition on each cell of image block to extract features. In this case I have bounding box for each car in the training images. In this situation I couldn't understand that how to get same dimension features for each bounding box. My bounding box size is not fixed but feature dimension. Could you help me more please?

[Image Analyst]

What are the features that you're measuring, once you've identified the bounding box that contains the car?

[FARHAD]

yeah. I have drawn bounding box that contains car only. In my image car as positive sample and others as negative. For different cars have different sizes box. Boxes are different size but I want same dimension feature. Another thing, I am using Contourlet Transform first time and so I don't know which features are useful for my work. I am very grateful to you. Please help me.

[Image Analyst]

Please post a picture. Also define dimension, as you are using it. If you have a 2D array of 5 rows by 8, is dimension (according to the way you are using it) "2" or "5 by 8"?