Similarity compatibility between images using multiple functions in Matlab

In a Matlab project for image extraction, I have extracted 14 functions from each image and shown in the following table.

** Si.No Feature name Size Example Values

**

  1      Feature_1      1x64    {96.02, 100.29, 69.04, 91.23,……89.42}
  2      Feature_2      1x64    {0.070, 0.0671, 0.0876, …….. 0.065}
  3      Feature_3      1x64    {0.837,  0.949,  0.992, 1.015 .…. 1.306}
  4      Feature_4      1x64    { 5.00,  5.831,  8.6023, 6.403,…..8.602}
  5      Feature_5      1x64    {-18.875,  -10.85,  -5.12,  … 39.2005}
  6      Feature_6      1x1      0.6465494
  7      Feature_7      1x1      0.89150039
  8      Feature_8      1x1      0.888859
  9      Feature_9      1x1      0.990652599
 10      Feature_10     1x1      157.8198719
 11      Feature_11     1x1      0.60112219
 12      Feature_12     1x1      0.060502114
 13      Feature_13     1x1      0.139164909
 14      Feature_14     1x1      5.7084825

      

        
        
        
      

    

The above feature set is for a single image. To calculate the similarity between two images, I tried the following methods.

First, I applied the distance calculation for a whole set of features by constructing a feature matrix (size: 14 xn) for the image.

Second, the 14 distance values ​​are calculated separately from the two images, after which the individual distance values ​​are added to obtain the final distance value. The problem here is that some features dominate and make other features ineffective. (for example, Feature_1 and Feature_10 give large distance values, so adding another 12 features, distance to them will have no effect. So I normalized 14 individual distance values ​​to a range of 0 to 1, again the problem with small distance values ​​becomes too much small after normalization.

But in both methods, the search results are not satisfactory. Are there any other methods for calculating similarity between images that account for all of the above features with equal importance?

Hello,

P.Arjun