A study on the relationship between depth map quality and stereoscopic image quality using upsampled depth maps - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 124

(1)
where D h and D u are ground-truth and upsampled depth maps, respectively.
2.1.2 Structural similarity index measure
A sophisticated tool for image quality evaluation is SSIM [ 14 ] that measures the similarity
between two images and considered to be correlated with the quality perception of the human
visual system (HVS). SSIM principle is based on the modeling any image distortion as a com-
bination of luminance distortion, contrast distortion, and loss of correlation. SSIM value for
two images f and g is expressed by
(2)
where l ( f , g ), c ( f , g ), and s ( f , g ) are luminance, contrast, and structure comparison functions, re-
spectively. σ f and σ g denote standard deviations, μ f and μ g are mean values and σ fg is covari-
ance. C 1 , C 2 , and C 3 are positive constants added to avoid a null denominator. The SSIM is a
value between 0 and 1 that higher value shows more similarity.
2.1.3 Visual information fidelity
Visual information fidelity (VIF) [ 15 ] is a full-reference image quality metric that uses inform-
ation theoretic criterion for image fidelity measurement. In an information-theoretic frame-
work, the information that could ideally be extracted by the brain from the reference image
and the loss of this information to the distortion are quantified in VIF method using natural
scene statistics (NSS), HVS, and an image distortion (channel) model. The VIF is derived from
a quantification of two mutual information quantities: the mutual information between the in-
put and the output of the HVS channel when no distortion channel is present (called the refer-
ence image information ) and the mutual information between the input of the distortion channel
and the output of the HVS channel for the test image. Similar to SSIM, the assessment result is
represented using a value between 0 and 1.