Enhanced image quality evaluation based on multi-feature for non-reference image quality evaluation

Reference documents
The main references in this section are: Leida, Shen, Jinjian, et al no-reference quality assessment of enhanced IMAGES[J]. China Communications: English, 2016 (9): 121-130.

Main ideas
In this paper, we mainly introduce the relevant ideas in this paper, which is to evaluate the quality scores of images after extracting the enhanced images.
This article mainly extracts five main features: Non-structural features-color continuity, color richness, contrast, illuminance and entropy. It is suggested that two other features-sharp
and natural characteristics, but these two characteristics in other papers have been introduced, it is not introduced.
Color continuity: For color images, first converted to HSI space image. For a pixel, if the hue (hue) is inconsistent with the hue of the surrounding pixels, it is considered
Pixel has an irregular hue. The formula is:
HABN is the number of all irregular tonal pixels, MN is the total number of image pixels.

Color richness: high-quality images with richer colors.
Contrast: Can be expressed by the histogram of the image, the richer the contrast, the more uniform the histogram. It feels like there's no innovation here.

Illumination: In fact, this method of calculation is the same as the mean value of the image and the median value.

which

Entropy: The higher the image quality, the greater the entropy.
A little Feeling
Although personally feel that this article in the big innovation point than other articles, but, very instructive, through the above five characteristics, we can put other methods to nest.