I am working on an image processing project where my first task is to classify a set of images as blurred or non blurred.
I am using only Python for implementation and the current status is as mentioned below.
I have generated the feature vector for a set of images (more or less 300 images) using color histogram and saved the feature vectors for each image line by line in a text file.
In this set of images there are two types.
Few images are perfectly fine ( images were captured well) so no need to process them. But there are images wherein some part of the image is affected by high intensity of light while capturing the image. However this set of images carry useful information as well. These images can't be discarded instead I need to remove the blurred region present in such images and use the remaining part of the image which carries useful information.
It is not well said in prior that which images are blurred and which images are not. I can check manually but its infeasible.
Could someone help me to classify blurred and non blurred images at the first step?
Thank you very much for your kind attention.
Try the gradient histogram span approach as a classification feature. In particular how blur and sharper images may have difference in edges that shows when this method is utilized.
I think you have firstly to minimize the histogram in order to get small range (normalized histogram) as I did in my paper titled "Improving Image Retrieval Through a Collection of Fast and Simple Visual Features Extraction" then you can use any classification methods.
@Abdulrahman Al-Molegi,
Thank you for your reply.
I have already normalised histogram values before writing them to a text file.
Only that I'm not aware which classifier suits the best. I'm a beginner in this field.
- Badges
- Science topic