Recently, I have proposed a new model for pattern recognition.
The model is examined on the faces images to build a face recognition system. It deals with continuous and discrete values of the extracted features.
The continuous Farhan model (CFM) represents each block of features with its mean and variance, whereas the discrete Farhan model (DFM) uses the quantization method to implement each block with a single value. The quantization method is also explained in the following articles:
The disadvantage of the quantization method is estimating the quantization levels and weights vectors that depend on the maximum and minimum values using a trial-and-error process. In the case of updating the database (add or remove an individual), the whole system is required to be retrained because the maximum and minimum values may be altered.
The question is:
How to assign a single value to each vector without using the quantization method. It is worth noting that each vector comprises only three elements; mean, maximum variance, and standard deviation.
Thank you in advance for your suggestions.