|Title||Mammogram image superresolution based on statistical moment analysis|
|Publication Type||Conference Paper|
|Year of Publication||2010|
|Authors||Wong, A., A. Mishra, D. A. Clausi, and P. Fieguth|
|Conference Name||7th Canadian Conference on Computer and Robot Vision|
|Publisher||IEEE Computer Society|
|Conference Location||Ottawa, Ontario, Canada|
A novel super resolution method for enhancing the resolution of mammogram images based on statistical moment analysis (SMA) has been designed and implemented. The proposed SMA method enables high resolution mammogram images to be produced at lower levels of radiation exposure to the patient. The SMA method takes advantage of the statistical characteristics of the underlying breast tissues being imaged to produce high resolution mammogram images with enhanced fine tissue details such that the presence of masses and micro calcifications can be more easily identified. In the SMA method, the super resolution problem is formulated as a constrained optimization problem using an adaptive third-order Markov prior model, and solved efficiently using a conjugate gradient approach. The priors are adapted based on the inter-pixel likelihoods of the first moment about zero (mean), second central moment (variance), and third and fourth standardized moments (skewness and kurtosis) from the low resolution images. Experimental results demonstrate the effectiveness of the SMA method at enhancing fine tissue details when compared to existing resolution enhancement methods.