Handwritten Tifinagh Character Recognition Through a Structural Approach

In this article, we present a handwritten Tifinagh character recognition system based on a structural approach which is graphs. It is composed of three main steps: pre-processing, features extraction and classification. In the preprocessing step, RGB thresholds, Morphological operations, normalization and thinning are applied to remove unwanted information from the image. In the feature extraction step, corners are extracted using a proposed algorithm then represented by an adjacency matrix defining the graph. These graphs are classified using a spectral graph matching method. Experimental results are obtained using multiple databases of handwritten Tifinagh characters to test the effectiveness. The system shows good results in terms of accuracy.