THE ADAPTIVE THRESHOLD BINARIZATION METHOD OF RASTER IMAGES OF TECHNICAL DRAWINGS

Abstract

The paper proposes a method of adaptive threshold binarization of technical drawings raster images. The specific image features of
technical drawings and their influence on the result of binarization of known universal methods are analysed. The artifacts that arise on the
resulting binary images are identified and ways to address them are suggested. The proposed method is based on the idea that very small parts of the image of the drawing resembles the letter «C», which in special cases degenerates into a straight line. Further arguments boil down to the choice of method for determining the conformity of the object in the image form «C». Separate effect given to the issues of binarization threshold settings depending on the brightness of the image. The paper presents a mathematical model of the proposed method, its algorithmic description and experimental qualitative and quantitative evidence of efficacy. The quantitative evaluation was performed by comparing the results with the reference binarization black and white image. As evaluation criteria considered criteria such as time of the algorithm, completeness and f-measure. The results of the conducted experiments demonstrating the superiority of the proposed methodology in the raster image binarization technical drawings, both quantitatively and qualitatively.