Methodology for classification of images based on the characteristics of a new interpretation of Shannon’s entropy

A new visual inspection method for the classification of wooden plates used in pencil manufacturing is presented. Darker regions in wooden plates indicate the presence of growth rings which are regions of hard wood. Pencils manufactured with these plates are more difficult to sharpen and have a tendency to bend and crack; therefore, these plates are classified as not being adequate for pencil manufacturing. The proposed method is based on the extraction and analysis of the features of the wooden plates using gray level images. The method classifies the plates using the results obtained by an automatic threshold determination based on Shannon’s entropy. The method was idealized aiming at low computational complexity, i.e., algorithm calculations involving only simple operations such as addition, subtraction, multiplication and division which could be implemented in hardware using VLSI technology or programmable logic devices. The wooden plate is mapped in an optimal number of regions; each region is pre-classified considering some relevant features based on the entropy gray level distribution of the pixels. Information from all regions is combined based on heuristic decision rules, arriving in a pre-classification stage where the regions are labeled in four classes (A, B, C and X). Two decision algorithms have been investigated for the final classification: one based in a co-occurrence matrix considering only a uni-directional horizontal neighborhood of the regions and the second is based on a heuristic method of information reduction considering combinations of the pre-classified regions. The final results obtained by the two algorithms were compared with the classification made by a human expert, demonstrating that the proposed method had very good performance.