A Block-Wise SVD Approach for Simultaneous ROI Preservation, Background Blurring, and Image Compression

This research introduces a novel approach to object detection and image processing, emphasizing precision in object detection and region of interest (ROI) isolation through block-wise Singular Value Decomposition (SVD). By leveraging the YOLOv8 model for object detection and integrating SVD-based transformations within non-ROI image regions, the method selectively processes desired subjects while compressing the surrounding background. The results demonstrate the effectiveness of the approach in enhancing object detection accuracy and achieving image compression. The block-wise SVD transformations exhibit versatility, allowing adaptation to specific application requirements. Key metrics, including compression values, size reduction, PSNR, SSIM, and RMSE, were considered for evaluation, revealing the method’s success in simultaneously enhancing object detection accuracy and compressing images. This innovative approach holds promise for various applications, particularly in fields like medical imaging, showcasing its potential impact on advancing image processing techniques.