I have some experimental images with different frequencies. After fusion I want to know how I can evaluate the quality of fused image. Any suggestions on some important parameter for same would be useful. To evaluate, I don’t have any reference image.
Whats your overall processing goal?
Do you use the image fusion as a part of pre-processing for subsequent tasks like feature extraction and classification?
If this is the case, it might be feasible to evaluate the quality of the fused images with respect to the overall processing goal. Use different fusion paramters and compare the impact on feature extraction and classification performance. If a certain parametrization of the fusion leads to better classification results (or vice a versa) you have a very intuitive quality measure regarding your specific processing goal.
If you are looking for more general approaches, the following work might give you a good starting point: 'A Survey on Different Approaches Used In Image Quality Assessment' - http://paper.ijcsns.org/07_book/201402/20140212.pdf
I hope that helps.
You can look at papers with the following keywords : No Reference Image Quality Assessment.
For example, one of the classic paper on this theme : "No-reference quality assessment using natural scene statistics: JPEG2000" by Sheikh, Bovik and Cormack ( http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=1518954 )
when the reference image is not available , the performance of fused image can be evaluated by using the following metrics such as standard deviation, entropy, cross entropy, fusion mutual information etc..
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing