Aleksandra Osowska-Kurczab
supervisor: Tomasz Markiewicz
Image resampling is frequently used as a preprocessing step in many computer vision tasks, such as classification or segmentation. Though image resizing is an inevitable step in many processing pipelines (e.g. utilizing pretrained architectures), tuning of the resizing method is usually omitted in the studies. Admittedly, there are many applications in which the exact influence of resampling on image textures and gradients is significant, e.g. medical image analysis.
The presented study analyses the impact of the image reconstruction in the downstream task of the renal tumour classification. A novel image reconstruction method is introduced, namely Sampling Kantorovich Algorithm (SKA). It is compared to six other popular techniques widely used in image processing. Based on the qualitative and quantitative analyses, it has been proved that the choice of image reconstruction method impacts the system's overall performance. SKA turns out to be the best performing method in the classification setup. It boosts the performance of the texture-based classifier to 75% of the weighted F1-score by approximately three percentage points (pp) compared to the best baseline solution. The results of this work may apply to a wide range of computer vision tasks, especially those established in medical image processing problems.