For simple implementation, Hopfield's neural network model is transformed into the equivalent model which does not contain inhibitory interconnection. In this case, thresholding level must be input-dependent unlike the original model. This equivalent model has been implemented to process 2-dimensional optical binary images. Neuron(logic) elements are represented by TNLC(twisted nematic liquid crystal) optical switch array and holographic interconnections are realized among these elements. Electrical devices are also used for thresholding operation. The implemented 16(4x4) neuron system behaved as content addressable associative memory, that is total image could be obtained from partial input which contains less than three bits error.