Муҳаммад ал хоразмий номидаги

Литература
 
1.
Хохлова Д. (2016). Бум нейросетей: Кто делает нейронные сети, 
зачем они нужны и сколько денег могут приносить,
2.
Molnar C. (2018). Interpretable machine learning.
3.
Olson R. (2016). TPOT: A Python tool for automating data science.
4.
Olson R.S., Moore J.H. (2016). TPOT: A tree-based pipeline 
optimization tool for automating machine learning.
IMAGE ENHANCEMENT USING CLAHE METHOD FOR THE IMPACT 
OF HYDROMETEORS
 
Suvanov Sh. M. 
Scientific-innovative center of information and communication technologies, 
Tashkent, Uzbekistan; [e-mail] 
Image enhancement plays a significant role in the enhancement of visual 
perception for computer vision, pattern recognition, and the handling of digital 
images. Histogram Equalization[1] is a mainstream method for improving picture 
contrast. Histogram Equalization is a standout amongst the most ordinarily used 
techniques in Contrast Enhancement since it has maximum efficiency and straight 
forwardness [2].
A modification of histogram equalization is the CLAHE (Contrast Limited 
Adaptive Histogram Equalization) method allows you to successfully process of 
images [3-4]. This method allows you to successfully process a fairly wide class of 
images. However, in some cases, when processing images distorted by 
hydrometeors (haze, fog, drizzle, rain, snow), this method produces images with 
the presence of artifacts and noise. It should be noted that these CLAHE features 
are show in the case of enough strong image distortions caused by hydrometeors, 
which affect different parts of the image with change degrees. 
In this paper, using CLAHE for color images has its features. First we have to 
convert an RGB image to HSV or CIELAB with luminance channel. Then apply 
the CLAHE method to this luminance channel. And finally, the image is returned 
in RGB format. 
This method algorithm is split the image into a set of tiles of the same size. 
Then calculate of the average brightness tile Y
mc 
and image average brightness 
Y
mgc
. Calculate for each color of the difference dY
mc
= Y
mc
- Y
mgc
. We recommend 
dY
mc
decrease. This can be done so that the objects in the image, having a 
brightness did not look dark. Calculation of the difference in brightness, where c = 
R, G, B. The calculation for each tile, taking into account the threshold P of the 

---

125 
histogram of brightness and cumulative brightness histogram for c dY. Using the 
cumulative distribution function obtained for each color, we calculate new RGB 
components. Then we receive the color image RGB. The proposed method for 
colour image as shown Figure 1. 
( a ) ( b ) ( c ) 
Fig 1. Images of the CLAHE method: (a) images distorted by hydrometeors, 
(b) using tile-average images, and (c) without using tile-average images. Threshold 
P=0.025. 
The images of Fig.1b often contain artifacts in the form of horizontal stripes. 
There are no such artifacts in the images of the Fig.1c. Despite the fact that all 
images were processed by the CLAHE method at the same threshold value P = 
0.025, the results of Fig. 1b noticeably differ from the images in Fig. 16c. 
Obviously, this is the result of pre-processing using the average brightness tile Y
m
. 
In this paper, the presence of hydrometeors in the atmosphere degrades the 
quality of the images, hides individual details of the image. One way to increase 
the visibility of objects in a distorted image is to adaptively align the contrast 
contrast histogram or CLAHE. This method allows emphasizing local contrast, 
which often increases the degree of distinguishability of image details. However, 
uneven hydrometeor distortion of the image can lead to artifacts in the image after 
applying CLAHE. The use of the proposed CLAHE algorithm for color distorted 
images allows us to compensate for possible artifacts arising from the application 
of the CLAHE method.