Image Quality Assessment - Science topic

ModelName % to open the model

print('-sModelName','-dpng','model.png')

This is a way to save ModelName.mdl in png format. Image name will be modemodel.png

Dear Hafijur Rahman, the trouble with human visual perception is that not all humans in fact have the same perception, which makes it hard to objectively quantify image quality. And we do need such objective technical comparison tools. Nevertheless, measurement tools for this purpose should be designed in such a way that they indeed match withe the average human observer. Even something relatively simple as the V-lambda curve that CIE defined to model the spectral sensitivity of the average adult human eye, does not match with all individual spectral sensitivities. Yet, this curve helps us to build comparable lux meters and design predictable lighting installations.

Both BRISQUE and NIQE are two well-known No-reference image quality measures in practice. You may note that No-reference image quality measures are different from reference-based image quality measures such as PSNR and SSIM, where we have the original image and another version of the image (may be degraded version).

Now, I will try to come o your question, the BRISQUE score from an image will be a positive value ranges from 0 to 100. A low BRISQUE measure indicates that the image is having high quality.

While considering NIQE measure also a smaller score indicates better perceptual quality. The range of possible NIQE measures are not defined.

If you have some experience in Matlab, you can try these two functions

S1=niqe(A);

S2=brisque(A);

then add some noise or apply filtering on the image A then check the same measures. You may get high values as compared the measures from the original image.

I hope the above-mentioned discussion will help you.

Dear Neil

To calcultad the MTF from edges or slits method you can use some available free softwares. I suggest the following:

(PDF) Free software for performing physical analysis of systems for digital radiography and mammography (DOI: 10.1118/1.4870955)

Best, Alessandra

Many parameters can be used such as; Peak Signal-to-Noise Ratio (PSNR), Root Mean Squire Error (RMSE), Standard Deviation (SD), smoothness, entropy, Structural Similarity (SSIM) can be found in

Don't think these are using SIFT features, but a number of python and matlab based implementation available in GIthub.

We (colleagues of mine)  used simple images collected by a camera. No EBSD maps! Unfortunately they gave it up several years ago. 

Image quality assessment metrics suggested by previous contributors are MSE and PSNR. These metrics are simple to calculate,mathematically convenient, and have clear physical meanings.  However these IQ assessments lack two critical features: 1) Ability to image similarity across stego image of varying distortion types; 2.) They are not very well matched to perceived visual quality. A new approach to assess image quality independent of the type and size of pixel size of concerned stego images is the estimation of Structural Similarity indices using  Structural Similarity Index metric(SSIM)

The author might need to see our investigation of SSIM computation of dissimilar stego images in secure e-voting application in this link:

Hi you can go through Ohta, N., Robertson, A.R., 2005. Colorimetry: fundamentals and applications. Wiley-IS&T series in imaging science and technology

Such methods exist, but they have different accuracy and may not reflect the various features of human vision. By the way, check Matlab Central.