Dynamic range independent image quality assessment

Abstract

Figure 1: Quality assessment of an LDR image (left), generated by tone-mapping the reference HDR (center) using Pattanaik's tone-mapping operator. Our metric detects loss of visible contrast (green) and contrast reversal (red), visualized as an in-context distortion map (right). AbstractThe diversity of display technologies and introduction of high dynamic range imagery introduces the necessity of comparing images of radically different dynamic ranges. Current quality assessment metrics are not suitable for this task, as they assume that both reference and test images have the same dynamic range. Image fidelity measures employed by a majority of current metrics, based on the difference of pixel intensity or contrast values between test and reference images, result in meaningless predictions if this assumption does not hold. We present a novel image quality metric capable of operating on an image pair where both images have arbitrary dynamic ranges. Our metric utilizes a model of the human visual system, and its central idea is a new definition of visible distortion based on the detection and classification of visible changes in the image structure. Our metric is carefully calibrated and its performance is validated through perceptual experiments. We demonstrate possible applications of our metric to the evaluation of direct and inverse tone mapping operators as well as the analysis of the image appearance on displays with various characteristics.

Dynamic range expansion based on image statistics

“…To find out which of these models yields better results we perform an expansion of the set of LDR images with the values obtained from Masia et al's linear model and those obtained with our multilinear model (Equation 5), and then evaluate the visual improvement of the expansion with respect to the original LDR image. For this purpose we use the image quality metric proposed by Aydin et al [3], which identifies visible distortions between two images independently of their respective dynamic ranges. The metric uses a model of the human visual system, and visible changes between a reference and a test image are classified into three types of structural changes: loss of visible contrast (when contrast visible in the reference image becomes invisible It can be seen that, the lower the key is, the more the two models tend to di↵er.…”

“…Fig. 7 Comparing the results with the image quality metric [3]. Expansion with the obtained in [17] and our , for the images Chapel and Frontier [13].…”

A Fast Semi-inverse Approach to Detect and Remove the Haze from a Single Image

“…In this work, we have employed the Image Quality Assessment (IQA) quality measure introduced recently by Aydin et al [20], which compares images with radically different dynamic ranges. This metric, carefully calibrated and validated through perceptual experiments, evaluates both the contrast and the structural changes yielded by tone mapping operators.…”

“…9 displays the comparative ratios of the (colored) pixels yielded by the IQA measure when applied to the results of the considered dehazing methods. Following the recommendation of Aydin et al [20], in order to reduce the possibility of misclassification, only the pixels with a probability scale higher than 70% have been considered. Based on the results from the table, it becomes clear that compared with the other techniques, the structural changes yielded by our algorithm are more closely related to sharpening operations (blue and red pixels) and less related with blurring (green pixels).…”

A Survey on Quantitative Assessment of the Quality of Tone-Mapped Images using TMQI-LI

“…The mainly addressed problems of HDR video compression and tone mapping that are important components of the high dynamic range imaging pipeline are discussed here [17]. In the HDR video part we proposed a new color space which encodes the full luminance range that is visible by the human eye.…”