Boundary Extraction of Tube-Type Bottle Based on Wavelet Restoration and Mathematical Morphology

In order to extract bottle boundary and pick out unqualified tube-type bottles which possibly have breaks and irregular shapes, a novel processing measure is taken based on wavelet transformation and mathematical morphology. Industrial bottle image contains noises so that traditional and theoretical image processing algorithms need to be improved. In addition, due to bottle inspection standard is strict in this project, appropriate edge image is the key factor for ultimate purpose. Consequently, in this paper, wavelet processing referred to decompositions in reasonable levels and restoration under rational condition is the first step to reconstruct image, eliminates major noises and simultaneously protects useful detail information in high-frequency part of bottle image. Because reconstructed image still has some noises escaping threshold control, and edge information possibly is damaged in conditional image-restoration, an improved mathematical morphological algorithm, including simple and multiple structure-elements designs, morphological repeating operations and arithmetical weighted calculation, has responsible for further de-noising, edge information compensation, boundary repairing and finally extracting high quality edge images assuring the whole pre-processing reaches to original purpose.