No-Reference Image Quality Assessment of Blur and AWGN Contaminated Images Utilizing Colour Features

Blur and additive white Gaussian noise are the two very common and dominant distortions in digital images. In this paper a No-reference image quality assessment (NR-IQA) technique is proposed which can estimate the quality of the images having these two distortion types covering entire subjective quality range. In the proposed method, derivative statistics and sharpness measurement are applied on the two colour spaces HSV (hue, saturation, value) and Lab ((Luminance, $a$ and $b$ colour channel) to extract the prediction features. The extracted features are fed to the support vector regressor (SVR) for training purpose. The experimental results shows the predicted quality score of the trained model are highly correlated to human visual system (HVS).