How do I extract and not just detect exactly one feature from a whole image?

Hi!Part of my project requires capturing a 2D real time image and extracting a car alone from it without any of the background details. I read up on CBIR and found a lot about image detection. I tried using SURF implementation also. I was getting the interested points, but retrieving the image back seems to be a challenge.The input image for our project will be dynamic and will keep changing every time so I am not able to fix a contour line even if I try to calculate the interested points based on that. I have attached the output i get from kmeans clustering based on colour. The only short coming is that this works only when there's a clear distinguishing color combination . car6 is the actual image and car6sample1 is the output. that is how i want the output to be even when there is no distinct color difference.Could you help me out with this?

[John BG]

why don't you try to extract the contours, then overlap the red you already got, and choose the contour that touches the red?

Do you already have the Image Processing Toolbox? In R2016 it is included the the basic pack.

There is this forum expert Image Analyst who may want to show you how to do it all

[Meera Girijan]

I'm using R2013. From which image should I extract the contours? Is it the original? I've got this red alone because I used a kmeans clustering method. If I have similar colors in the background or for that sake different colors within the car structure I might not get such a clean image. in that case what do I do?

Answer 1

[Walter Roberson]

Well, if you are going to insist on clustering on color, then add a picture of a yellow banana and a red plantain to your image before clustering, so that you can be sure that you have contrasting colors so the algorithm will be sure of finding something. The result will be about as useless as clustering by color, but will have the assurance that at least two clusters exist in the image, which you otherwise could not be certain of.

[Image Analyst]

Meera, Walter is right about kmeans being not a good method for finding cars, and also about being forced to find the number of clusters that you specify regardless of how many distinct clusters there actually are in the image.

A car is not one feature like you asserted - it's a whole bunch of features. Soon Walter will report his post showing all kinds of things that could be considered cars but are barely recognizable as such. You need to do further study on CBIR and image understanding. Sorry, this is not my main field of research.

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/264009-i-need-a-coding-to-detect-the-cars-in-an-images#answer_206408