I applied three techniques for contrast enhacement on the image "less Contrast.jpg", so i obtained three different images : "MSRCR1.jpg, IlluEnhHSV1.jpg and IlluEnhHSVMSRCR1.jpg", I would like to know what is the best one from these three images (compared to the orignal one) in terms of Hue, Saturation, Value, clarty of edges and the noise ?
Best Regards
Khadidja BELATTAR
I mean that best in terms of Hue, Saturation, Value, clarty of edges and the noise.
The best image will be used for the melanoma recogntion using color and edge features.
Yes, that's my question. There is nothing like "best in this and that in general". Well, you can use some noise estimators and compare (like RMSE, PSNR, ...) but anyway - I would suggest you to try all of your outputs to your concrete application and you will see what gives you best results.
For visual assessment I personally would prefer MSRCR1.jpg.
Though there seems to be little difference between this and IlluEnhHSVMSRCR1.jpg while IlluEnhHSV1.jpg is clearly inferior.
BTW: 'contrast enhancement' should not affect 'noise' (but is capable to make it more visible). More or less the same applies for 'clarity of edges'.
As of my concern , i personally suggest MSRCR1.jpg..(edges are more clear as compared to other images)
There are a lot of objective and subjective methods. You can read the paper
A Survey Over Image Quality Analysis Techniques for Brain MR Images, Int Journal of Radiology 2015 June 2(1): 29-37, doi:10.17554/j.issn.2313-3406.2015.02.5
As the application is melanoma recognition and you specified the desired attributes are colore and edge information then you can use structural similarity index as SSIM. It is good for quantification rating of your algorithm but in the same time see the qualitative check of processed image.
there are non-reference quality measures such as brisque, niqe, cqa etc that can guve you image quality measurement without the need for original source image
Hi Atif Shah,
You can find melanama images on these links:
http://www.fc.up.pt/addi/ph2%20database.html
https://licensing.eri.ed.ac.uk/i/software/dermofit-image-library.html
Best Regards
You are welcome Atif Shah,you must pay £50 for licence fee.
Good Luck
Best Regards
Khadidja BELATTAR
Thank You dear Wolfgang Reichl ,
I would like to use the best preprocessed images for diagnosis . I mean the best images in terms of hue, saturation and brightness. Because the color criterion is considered as a measure for diagnosis melanoma image.
Best Regards
Khadidja BELATTAR
- Badges
- Science topic
- Similar topics
- Computer Science
- Images