Vector quantization (VQ) is a pattern classification method in which each set of vectors represents a particular class or category. It is also can be defined as a mapping of input vector into a finite collection of predetermined codevectors. The set of codevectors is called codebook. The mapping should be performed so as to minimize the loss of useful information for a particular application. There are many algorithms and techniques that are used for VQ such as classical clustering techniques (such as the k-means). Recently, neural Network is used as another approach to perform VQ. The work presented is an attempt to use the Higher order and Quantized Hopfield neural Network for vector quantization, and study whether these two approaches give a better results. A High order version of the Hopfield neural Network is presented to perform a simple vector quantization or clustering function. This model requires no penalty terms to impose constraints in the Hopfield energy, in contrast to the usual one where the energy involves only quadratic terms in the state vector. The main problem that the researcher faces when using the standard Hopfield net is the difficulty of choosing appropriate constant values (that are multiplied by the penalty terms). In this work, such a problem leads to the occurrence of odd blocks (those blocks that are not mapped by any codebook index, or mapped by more than one codebook index). To get rid of these odd blocks, each of them was mapped to its closest codebook vector. In this work, Quantized Hopfield Network (which was suggested by S. Mastuda to solve integer programming [mat99]) was implemented and used for image quantization and proved successful. The neurons in this net. called quantized neurons, which takes quantized values such as integer values, rather than just binary values or continuous values (as with traditional Hopfield nets). In this model, number of neurons, connection and computation time can be decreases greatly as compared to the traditional Hopfield networks. Therefore, this net is faster than the other traditional Hopfield nets. It obtains the best quantization result compared with classical methods and traditional Hopfield Nets. The main problem with Quantized Hopfield net is that, its performance is affected by the codebook initialization To solve the problems that may occur with the used methods, a new method was suggested which is a combination of the Standard and Quantized Hopfield. The suggested method makes use of the benefit of both Standard Hopfield (provides automatic codebook initialization), and Quantized Hopfield (to get rid of odd blocks). Finally, a comparison was made between various method (k-means, Higher order. Standard and Quantized Hopfield) to study the behavior of these methods in solving vector quantization problem. The system was implemented using visual Pascal programming language (Delphi version 4.0) on personal computer (Pentium II processor with 32 MB RAM and 4GB hard disk that works under Windows 98).