ABSTRACT
Communication technology and information systems play a vital role in medical science. Telemedicine applications use communication and information technology to provide healthcare services and enable transferring patient information between distance locations. Medical image security and authorization is an essential component of telemedicine application that ensures the privacy and confidentiality of patient information. During acquisition, coding, authorization, and the encryption process, the image quality deteriorates. Image quality assessment is an important aspect of medical imaging to ensure image quality for accurate disease diagnosis and treatment. Image quality assessment (IQA) methods are useful for image security and authentication by detecting whether an image has been tampered or altered. A No-Reference Image Quality Assessment (NR-IQA) method is proposed in this chapter for image security and authorization without the need of a reference image. The proposed methodology works on the hypothesis that distortion present in an image changes the naturalness of the image features. The unnaturalness of the medical images is found in wavelet domain by fitting the sub-band coefficients distribution using a Normal Inverse Gaussian (NIG) Probability Density Function (PDF). The noise feature is calculated by fitting a Gaussian PDF in diagonal sub-band. Further, the Natural Scene Statistics (NSS) and spatial correlations features are calculated. Support Vector Machine Regression (SVR) is implemented for classification and regression to estimate the quality factor. The experimental results show the performance improvements of the proposed method are 0.84%, 1.12%, and 5.04% over the state-of-the-art NR-IQA methods in terms of SROCC, PLCC, and RMSE parameters, respectively, on various distortions including JPEG 2000 (JP2K), JPEG, additive white noise (WN), Gaussian Blur (GB), and Fast Fading (FF).
