<title>Statistics of visual representation</title>

Abstract

The experience of retinex image processing has prompted us to reconsider fundamental aspects of imaging and image processing. Foremost is the idea that a good visual representation requires a non-linear transformation of the recorded (approximately linear) image data. Further, this transformation appears to converge on a specific distribution. Here we investigate the connection between numerical and visual phenomena. Specifically the questions explored are: (1) Is there a well-defined consistent statistical character associated with good visual representations? (2) Does there exist an ideal visual image? And (3) what are its statistical properties?

Hardware Implementation of an Automatic Rendering Tone Mapping Algorithm for a Wide Dynamic Range Display

“…Previous studies on perceived naturalness and image quality have shown that the observed naturalness of an image correlates strongly to the characteristics of the image, primarily brightness, contrast and image details [41][42][43]. As previously highlighted, the factor k is a major parameter in the tone mapping algorithm, and its value may vary for different WDR images.…”

“…The size of the pixel regions for computing the standard deviation can be modified. Based on the results from their experiment, they concluded that a visually optimal image was found to have high image contrast, which lies within 40-80, while its image luminance is within 100-200 for an eight-bit image or, more specifically, around the midpoint of the images dynamic range [41]. This is to ensure that the image has a well-stretched histogram.…”

“…The overall image luminance is measured by calculating the mean of the image's luminance, µ, while the contrast σ is measured by calculating the global mean of the regional standard deviations. To obtain the regional standard deviations, non-overlapping area blocks of 50 × 50 pixel size were used to calculate the standard deviation of a pixel region in their experiment [41]. The size of the pixel regions for computing the standard deviation can be modified.…”

“…The statistical method by Jobson et al [41] was used in estimating what the illumination range should be in order for an image to be perceived as being a visually pleasant image. In Jobson et al's work, they developed an approach of relating a visually pleasant image to a visually optimal region in terms of its image statistics, i.e., image luminance and contrast.…”

Color image with Dim regions Enhancement Using Modified Histogram Equalization Algorithm

“…Some researchers in NASA Langley research center [13] concluded heuristics to measure quality of the image, proposed the idea that good visual representations seem to be based upon some combination of high regional visual lightness and contrast. To compute the regional parameters, we divide the image into non overlapping blocks that are 50×50 pixels.…”