Our software developer Aleksi Yrttiaho has finalized his thesis ‘Inpaint Mehtod for X-.Ray Image Restoration’. Thesis focuses on refining digital raw X-ray images by removing defects through post-processing techniques. Post-processing of raw images is a key research and development focus for Innomentarium Oy in enhancing its in-house medical imaging workstation software.
X-ray detectors have defects from the factory, which appear as white pixels, usually in groups of dots or lines, in the output raw images. These areas obscure parts of the X-ray image, interfering with interpretation and analysis. Therefore, these defects must be removed. The objective is to restore the image by eliminating defected areas so that they become indistinguishable to the bare eye, resulting in a natural, research-quality X-ray image. The thesis describes the developing process for a suitable image inpainting algorithm to restore X-ray images.
Before starting the development of the restoration application, the topic was explored through a literature review. The aim of the research process was to understand the principles of X-ray imaging and identify suitable image inpainting methods for restoring X-ray images. Subsequently, the programming phase began, during which the algorithm was designed and implemented. Following this, algorithmic tests were conducted to examine the functionality of the application and the quality of the restored images. Finally, an expert tester conducted a comparative analysis using a set of images, half original and half restored, to evaluate how authentic and high-quality the restored images appeared to the human eye.
Based on the algorithmic results, the restoration algorithm integrated into an image capturing application, proved functional and capable of producing high-quality restored X-ray images. The test involving the test person produced positive outcomes, as they did not discern a difference between the original and restored images. While the quality of the restoration may decline if the areas to be restored in the image are too large, the algorithm has proven effective for restoring smaller image areas.
