Evaluating Resizing Techniques for Lumpy Skin Diseased Animal Image using Metrics

In medical imaging, image preprocessing is especially significant in the analysis of lumpy skin disease in animals. These challenges are well-marked, from preserving quality during resizing across different algorithms used with those techniques in a dataset. In the paper, several recent resizing methods are discussed including fixed size, aspect ratio preservation, crop, and even more advanced methods of interpolation; these include nearest neighbor, bilinear, bicubic, and Lanczos interpolation. These interpolation techniques were applied on images of infected cattle skin in order to focus on performance-based metrics such as Structural Similarity Index (SSIM), Mean Squared Error (MSE), and Peak Signal-to-Noise Ratio (PSNR). The proposed objective is to evaluate whether these resizing techniques preserve the quality of the images. The results show that advanced interpolation methods, such as bicubic and Lanczos, are particularly good at achieving a high similarity with the original image. However, the analysis was limited to a single image, so validating the conclusion across a more diverse set of datasets is necessary. Future work involves combining recent resizing methods with the extension of this approach for images with other conditions, such as blur, brightness, and contrast variations, to improve the generalizability and robustness of such methods.