# calculate image distortion level

30 views (last 30 days)
Sunetra on 22 Jul 2024 at 5:53
Commented: Sunetra on 23 Jul 2024 at 6:02
I have nine similar images, each exhibiting different qualities of visual patterns. Each image is composed of a single unit that is repeated multiple times, either randomly or in a uniform pattern, to form a complete image. My objective is to assess the extent of disruption each image causes to the human visual system. Unfortunately, I do not have a reference image for comparison.
I have attempted various methods to analyze these images, including mean geometric disruption, blur disruption, color disruption, and ranking the images based on quality. However, none of these approaches have provided a satisfactory analysis. I need an idea to find a metric to calculate the distortion level of each image to the human visual system.

Shubham on 22 Jul 2024 at 7:37
Hi Sunetra,
To assess the extent of disruption each image causes to the human visual system, you can consider using a combination of metrics that capture different aspects of visual perception. One approach you might find useful is to use a combination of structural similarity index (SSIM), entropy, and frequency domain analysis. These metrics can help you quantify the level of disruption in each image.
Here's a step-by-step guide on how you can implement this in MATLAB:
1. Structural Similarity Index (SSIM): SSIM is a method for measuring the similarity between two images. It considers changes in structural information, luminance, and contrast. For your case, you can compare each image to a reference image or a uniform pattern. Refer to this documentation: https://in.mathworks.com/help/images/ref/ssim.html
2. Entropy: It measures the amount of information or randomness in an image. Higher entropy indicates more complexity and potential disruption. Refer to this documentation: https://in.mathworks.com/help/images/ref/entropy.html
3. Frequency Domain Analysis: Using Fourier Transform, analyze the frequency components of the image. High-frequency components often correspond to edges and fine details, which might contribute to visual disruption. Refer to this documentation: https://in.mathworks.com/help/signal/ug/practical-introduction-to-frequency-domain-analysis.html
Combining these metrics by normalizing each one and summing them up to create a composite score. Rank the images based on these composite scores to determine the extent of disruption.
I hope this helps!
Sunetra on 23 Jul 2024 at 6:02
Thank you for the response. How to check which feature is making the images more disruptive than the others?