The Study On Image Enhancement Algorithms Of Infrared Thermal Imager

Infrared thermal imager is a device which can convert infrared radiation from the surface of the object into visible image; infrared thermal image is the final form of expression. Infrared thermal image shows the temperature distribution of infrared radiation from target surface which is measured. High sensitivity, accurate temperature measurement and high reliability are the natures of infrared thermal imager. Therefore, in recent years, infrared imaging system has been widely used in the military, medical, industrial and agricultural fields. In the meantime, higher image quality of thermal imaging systems has been required by people. However, due to the reasons of the mechanism of the infrared thermal imaging system and the infrared thermal imager itself, compared to the visible images, infrared images have the following characteristics: low contrast, more noise, blur edges, and the boundaries of temperature range is not obvious. These features are unfavorable to the following infrared image processing, on the other hand, if using Human-computer interaction, preprocessing of the original infrared image is necessary to improve the visual effects, making the image more suitable for further analysis and processing. Therefore, it has the necessity and urgency on infrared image processing.Uncooled infrared imaging system has advantages of low prices, high reliability, small size and low power consumption, so it occupies an important position in the development of infrared thermal imaging system. In this paper, processing the images obtained from 320×240 uncooled infrared focal plane infrared thermal imager ,the main achievement is following: low contrast, more noise and blur edges are characters of infrared image, the author attempts to study infrared image enhancement through the perspective of studying infrared image contrast enhancement, noise reduction and infrared image edge enhancement. In addition to the classical image enhancement algorithm, Retinex algorithm based on the theory of visual constancy has been introducedin infrared image processing, and achieved significant results. The results of this paper are as follows:According to the producing principle, features of infrared image are analyzed, and histogram, features of noise, contrast and resolution are studied deeply. We know infrared image is generally dark and the goal between backgrounds is low contrast, blur edges, and more noise. So the author attempts to study infrared image enhancement through the perspective of studying infrared image contrast enhancement, noise reduction and infrared image edge enhancement.The classic image enhancement algorithm of infrared images is studied in this paper. From different perspectives, the author used different methods of image enhancement. After carrying out a large number of simulation experiments and reached the following conclusion: histogram equalization is most effective in image enhancement and the contrast of the image after enhancement is greatly enhanced. This method is simple and easy to implement, but there are excessive enhancement in some images, the image after processing is often harshness and not soft enough. In the fact of noise reduction, the effect of median filter in random noise (Slat& peeper noise) is better than Gaussian noise. It can protect the image edge very well when reducing noise; it is an ideal non-linear filter. For image edge enhancement, Laplacian algorithm is better than Sobel algorithm. In frequency domain, Butterworth filter upgrade to high-frequency has the most notable improvement. It retains the effect of high pass filter to the image sharpening, and also to enhance the gray scale of the overall image, and making the Infrared images noticeably improved.Finally, the author introduced the Retinex algorithm which based on visual color constancy to infrared image processing. The purpose of Retinex theory is to obtain the reflectivity of the object from the image which is formed by reflected light. That is to say, it obtains the original appearance ofimage not considering the illumination. But the mechanism of infrared image is different from the visible light images. We assume that the infrared radiation is a reflection of the objects which are in the infrared light source, and the infrared image is formed by the infrared lights. This is in line with Retinex theory. Based on this assume, in this paper, we used Retinex algorithms in infrared image processing and analyzed the effects of infrared image which processed in different Retinex algorithms. The effects of infrared images which are processed by Frankle algorithm are better, because we can see the image details and contrast are enhanced, but some images have appeared Halo which is the weaknesses of Frankle algorithm. In the center/around Retinex algorithm, we discussed different parameters of SSR and MSR algorithm and the effects of infrared image processing. At last, we do a treatment to the output image, and achieved good results.In this paper, the infrared image enhancement algorithms, which studied by author are simple and fast. So this paper will play a guiding role in enhancing the image quality of infrared thermal imager.