ADAPTIVE DIGITAL PROCESSING OF VETERINARY RADIOGRAPHS FOR COMPUTER VISION TASKS

Author(s): Titov A.D.

Rubric: Information technology

DOI: 10.21777/2500-2112-2026-3-97-108

Release: 2026-3 (56)

Pages: 97-108

Keywords: veterinary radiology, vertebral heart score, image quality enhancement, digital processing, histogram adaptation, morphological processing, computer vision

Annotation: The relevance of this study stems from the need to improve the accuracy and reproducibility of automated vertebral heart score (VHS) calculation in small companion animals. One of the key factors affecting the quality of anatomical structure segmentation is the low contrast of original radiographs, particularly in the regions of the thoracic vertebrae and the cardiac silhouette. This article proposes an adaptive digital image processing algorithm that combines local contrast enhancement, morphological extraction of bony structures, and nonlinear edge-preserving noise suppression. The algorithm is designed for use at two stages of the model lifecycle: during dataset preparation for expert annotation, and as a preprocessing step prior to inference of a trained computer vision model. The method description includes mathematical formalization of the components, justification for parameter selection, and the architecture of the algorithmic pipeline. Experimental validation of the developed algorithm’s effectiveness was conducted, including a comparison of annotation quality and the performance of models trained on original versus processed images.

Bibliography: Titov A.D. ADAPTIVE DIGITAL PROCESSING OF VETERINARY RADIOGRAPHS FOR COMPUTER VISION TASKS // Education Resources and Technologies. – 2026. – № 3 (56). – С. 97-108. doi: 10.21777/2500-2112-2026-3-97-108