An improved unsharp masking model for intensive care unit chest radiograph enhancement

Portable chest radiography (CXRs) now widely used in intensive care unit (ICU) and performed on the patients who need best diagnostic tools. Nevertheless, these CXRs are usually in poor quality, mainly due to exposure settings and the misaligned body position. To help clinicians to detect the subtle pathological changes, we need to enhance ICU CXRs. Un-sharp masking model is a very powerful tool to enhance image. Generally, it uses Gaussian filter to smooth the image. However, this method suffers from deficiency. Gaussian filter smoothes evenly both the edges and other fine details, which will cause amplification of noise in homogeneous areas and details areas and other problems. For the purpose of overcoming the weakness of classical un-sharp masking model, we improved this technique by replacing Gaussian filter with anisotropic diffusion filter. Compared to the traditional Gaussian-filter-based un-sharp masking method, our technique sharpens the edge while the noise, on the other hand, is barely enhanced.