Combined transform and spatial domain based “no reference” measure for underwater images

Underwater images suffer from light-induced degradation, since light rapidly decays as it moves through water. They also incur blurring effects, loss of color or greyed out images due to scattering, low contrast due to absorption. Even though there are many image enhancement algorithms for enhancing underwater images, there is a lack of quality measurement techniques that can quantify the quality of enhancement. Despite the development of several image quality metrics, there is still a need for a single metric that can extract the most perpetual quality of the image. The existing state-of-art quality metrics can be grouped into two large groups: image quality assessment in the spatial domain (ICA, IME) and image quality assessment in the transform domain (Fourier, cosine, discrete wavelet transform). In this paper, a new approach is presented for measuring the quality based on the combination of the transform domain and the spatial domain (pixel intensity) approaches. At the heart of this approach, is the pyramidal decomposition of images into multiple scales and assessing the quality of each stage. The overall image quality is then evaluated using human visual system based combinations of qualities at each stage. The experimental results demonstrate that the measure effectively evaluates the quality of underwater images and is in close accordance with human visual perceptions. Underwater image quality measurement has practical applications in the analysis of aircraft or watercraft inspection and visual hull inspection.