This paper presents an implementation of hand gesture recognition by histogram and matching approach. This implementation is applicable to the human hand, just an object of ROI of very little texture. The method of our implementation has three steps. First, considering a pre-take base image as our gesture model, or by retrieval from the feature database of the training hand gestures, the base and candidate images are converted to two-dimension histograms respectively. Second, four statistical factors are computed. In the last step, a similarity value will be measured of each candidate image based on the factors we computed.