A Hybrid Algorithm for Contour Thinning in Image Processing
Abstract
Visual data processing is a rapidly expanding field, with image processing aimed at enhancing image features for object recognition. Typically, an initial captured image undergoes pre-processing, followed by segmentation. Segmentation results directly depend on the results of image processing stage. Precise contour separation of objects is crucial for effective segmentation; however, contouring methods often produce thick contour lines, leading to unnecessary information in the final image. Contour thinning algorithms address this issue, yet many existing algorithms lack real-time applicability and fail to deliver precise contour separation. The use of modern deep learning algorithms is considered inefficient due to the fact that they require large amounts of training data, a large number of computing resources, and are not adapted to work in real time. This highlights the need for advanced algorithms that meet real-time requirements and improve contour accuracy. In this research work, a hybrid algorithm for thinning image contours that has passed through image pre-processing steps is proposed, and it is compared with existing algorithms in terms of pixel matching and time criteria. Computational experiments demonstrate that the hybrid algorithm outperforms current options, making it a practical recommendation for real-time applications.