Hi,
I would like to investigate the effect of the preprocessing images on image retrieval tasks. So I'm thinking about preprocessing the images than I compute the accuracy of the retrieval after and before preporcessing. But, it seems to me to not be a good idea.
Another idea, is to preprocess the images, instead of computing accuracy of retrieval, I compute the segmentation rate of the preprocessed images. But my problem is I don't have the ground truth (i.e segmented images by the experts) of the used images.
What do you think?
Thank You.
Best Regards.
Khadidja Belattar.
I think you should calculate and compare 2 retrieval accuracies; one without preprocessing and the other with preprocessing.
I think the question is broad: What specific retrieval task you want to accomplish? What specific pre-processing steps you want to implement to hopeful improve the retrieval task? With these questions answered, we may be able to suggest the effects of pre-processing on retrieval tasks.
Hi Baraka Maiseli ,
I would like to retrieve similar skin cancer images for diagnosis task so, to do that i must firstly preprocess the skin cancer images;removes hair,air bubbles, ruler murking shadow and improve contrast.
Thank you Baraka Maiseli .
From your application point-of-view, I would go with Nilanjan Dey and would advise going with IQA study. However, since you don't have a ground-truth to compare with, you probably require a blind image quality analysis where the reference image is unavailable.
A good paper on the same is here:
http://www.kresttechnology.com/krest-academic-projects/krest-mtech-projects/ECE/dspmt/%5B25%5D.pdf
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts