Applications of Wavelet Coefficient Estimation in Medical Image Enhancement

• Main Authors: Wen-Li Lee, 李文禮
• Other Authors: Mei-Juan Chen
• Format: Others
• Language: en_US
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/19027423592773343441

博士 === 國立東華大學 === 電機工程學系 === 99 === Medical images provide clinical information for facilitating diagnostic accuracy and treatment process. Clear image detail is essential and can provide better information for visualization. When the information in medical images is not satisfactory to viewers, image enhancement can be used to improve visual perception on the images. However, over-enhancement or under-enhancement could happen in some images. Moreover, the information obtained via visual inspection on the enhanced images can be varied by each viewer because the preference of visual perception is individualized. It would be beneficial for clinical practice if the viewers can select the optimal enhanced images for their medical use. We proposed three wavelet-based methods specifically to improve visibility on digitized medical images in terms of resolution enhancement, detail enhancement and texture enhancement. Wavelet coefficient estimation is the core technique employed in these three methods. Our proposed wavelet-based interpolation method enables arbitrary resizing medical images and reduces influence of image blurring. Our proposed detail enhancement scheme can sharpen image and reveal hidden information, so visibility on medical images can be improved. We also successfully integrate these two methods, wavelet-based interpolation and detail enhancement scheme, to achieve resolution enhancement and detail enhancement simultaneously. Furthermore, we proposed texture enhancement scheme to increase definition of texture in noise-corrupted sonograms without eliminating speckles. Experimental results show that our proposed methods outperform other schemes commonly used for medical image enhancement in terms of subjective assessments and objective evaluations. Our proposed methods allow scalable selection of level of image enhancement to meet viewers’ visual preferences.