Contrast enhancement of infrared images using Adaptive Histogram Equalization (AHE) with Contrast Limited Adaptive Histogram Equalization (CLAHE)

Abstract


Introduction
The digital image processing can be defined as the set of techniques applied to digital images in order to improve quality of images generally and Infrared ones especially [1]. Infrared systems provide thermal information, which is invisible for human perception, as well as some limitations. This is because that thermal radiation from an object is affected from scattering due to the atmospheric conditions [2].
Infrared radiation (IR), sometimes referred to infrared, is a region of the electromagnetic radiation spectrum that wavelengths range from about 700 nanometers (nm) to 1 (mm) [3,4]. (IR) images are typically degraded due to noise, low contrast (since IR image sensors cannot clearly distinguish objects from their backgrounds if they have a similar emissivity) and blur (due to the inhomogeneous photosensitive response of the infrared detector and non-ideal optics system) [5].
Infrared light is invisible to the human eye, even though longer infrared waves can be sensed as heat. Warm objects emit infrared energy and the hotter the object, the shorter the wavelength of IR energy emitted. This IR emission enables rescue workers equipped with long-wave IR sensors to locate a lost person in a deep forest in total darkness.
These wavelengths include most of the thermal radiation emitted by objects near room temperature. Images received through various infrared (IR) devices in many applications are distorted due to the atmospheric aberration mainly because of atmospheric variations and aerosol turbulence [6,7].
In this paper the zenmuse XT thermal camera will be used with quadcopter MATRICE 100. Type As with other DJI systems, a 3-axis gimbal systems, the zenmuse XT stream alive HD view to DJI GO software.

Contrast enhancement
The main standard aims of contrast enhancement of infrared (IR) images are: 1-To make them very effective in an application.
2-Provide more appealing image, with easier differentiation of objects. Contrast operator is one of the factors of low or perfect quality images. Because of an image cannot be said to be of good quality when it has very low contrast or too high contrast [8]. Contrast image enhancement techniques divided into three different categories [9]: Global enhancement, Local enhancement and Adaptive Enhancement.
Histogram equalization (HE) is a widely used global contrast enhancement technique for color and grayscale images together. The histogram equalization (HE) is a method to obtain a unique input to the output contrast transfer function. HE spreads out and flattens the histogram of the number of image pixels at each gray level value [10]. The main idea in adaptive histogram equalization (AHE) is to take into account histogram distribution over the local window and combine it with global histogram distribution.

The proposed technique
The proposed algorithm combines these two methods (AHE&CLAHE) and applied it on an images and compare these results with default algorithm.
The first step is dividing the image into several non over lapping regions of approximately equal sizes. The second step is calculating the histogram of each region. Then, based on the desired limit, obtain a clip limit for clipping histograms for contrast expansion. Next, redistributed each histogram in such a way that its height does not go beyond the clip limit.
The clip limit β is obtained by: [11,12] (1+ where α is a clip factor, if clip factor is equal to zero the clip limit becomes exactly equal to ( ), furthermore if clip limit is equal to 100 the maximum allowable slope is . Finally, determined the cumulative distribution functions (CDF) of the resultant contrast limited histograms for grayscale mapping. The pixels are mapped by linearly combining the results from the mappings of the four nearest regions.
The proposed algorithm which combines the two methods (AHE) & (CLAHE) can be summarized as the following steps: Step1: read the image.
Step3: enter a matrix of image size.
Step4: define the variable that will use to display the histogram image.
Step5: calculate the histogram of the image in the range [0, L-1] by using the discrete function: ) = Step6: plot the histogram distribution curve.
Step7: calculate the cumulative density function for the histogram that given by: Step10: normalize the image of histogram equalization Step11: calculate the new distribution of the image histogram.
Step12: divide the image to three regions according to specification of every region that different from another region.
Step13: calculate the slop of every region. Step14: calculate the clip limit by using the equation: (1+ where α is a clip factor that different from region to another, and can be determined according to the characteristic for each region.
Step15: Clip the histogram above the clip limit and use it to find CDF.
Step16: Now map this intensity into the output image with range [min max].
Step17: calculate the AHE for the image using MATLAB function and make comparing between this result and the result from our algorithm.

The experimental results and discussion
Areal images (vertical and horizontal thermal images) have been used in this research. These images have been taken from different heights. Some images were in size of (512x640) and the other of size (480x720), in addition to a different amount of noise. Different enhancement techniques were used in order to determine which method is the best for thermal image.
This algorithm was applied to all vertical and horizontal images. The results of this algorithm can be shown in Figs.1-4.

Conclusions
It is obvious when applying the proposed algorithm on the vertical and horizontal thermal images, there is a clearly changes in the images comparing with the original images in intensity distribution and as well as the histogram of the image was changed too and became more uniform, the edges became more sharpening.
The image that doesn't have too many variations in the levels of gray image was showed that AHE has not good results. To overcome this problem, the noise is suppressing by contrast limited enhancement and the result images is better. One important thing that AHE has the disadvantage that AHE enhances not only the image, but also it enhances the noise of the image, while this method enhances the image and suppresses the noise.