Efficient Image Enhancement Algorithm Research Based On Histogram

Image enhancement is an important part of image processing, its objective is to emphasizesome prominent certain features and suppress background or noise by improving the image contrast,and improving the visual effects and quality of image. According to the different processing spaces,image enhancement can be categorized into two types, spatial domain based and frequency domainbased. Although the frequency domain based algorithm can enhance the contrast effectively, thespatial domain based algorithm is applied more widely. As frequency domain has complicatedcomputation that is not suitable for real-time processing.Histogram equalization (HE) is the most popular algorithm in spatial domain because of itssimple operation and significant effect. It is widely used in the medical image enhancement. But HEhas a serious problem that is "mean shift". When the low dynamic range image is processing, it willlead to excessive enhancement and produce artifacts and edge effect. So it is difficult to applydirectly to the TV and other consumer electronic products. But with high demand of brightnesspreserving in the consumer electronic products, we urgent need an efficient brightness preservinghistogram equalization algorithm to solve it. In addition, the traditional image enhancementalgorithm based on histogram can effective enhance the low contrast images of small range ofdynamic, but not suitable for the images of big range dynamic. Retinex theory can separate the truecolor of objects from the original image. It can remove the effect of illumination on the object, andapplicable to all kinds of contrast image. This paper make research respectively on the traditionalhistogram equalization image enhancement and Retinex theory, some details as below:Firstly, this paper elaborates the basic principle of image enhancement, and introduces threemethods of the spatial image enhancement: direct gray-scale transformation, histogram processingand spatial filtering. Frequency domain image enhancement mainly include all kinds of filters, suchas low pass filter, high pass filter, homomorphic filter and selective filter. At the same time, weintroduce the color models in common use in detail, such as the device oriented RGB and CMYcolor model, YUV/YIQ and HSI which are more in accordance with the human vision model.Secondly, this paper proposes a new brightness preserving image enhancement algorithmbased on constant histogram model, and introduces the proposed algorithm through the basicprinciple, the model establishment and derivation and the experimental results in detail. The essenceof this algorithm is the histogram specification, so we make a further introduction of histogramspecification and make a comparisons of some commonly used histogram specification algorithm. We compare the results of our proposed algorithm with some classic algorithms by choosing threedifferent image quality measurement standard, and confirmed that the proposed algorithm has highperformance on the computational complexity, brightness preserving and contrast enhancement.Finally, this paper also do some research on Retinex theory. We elaborate on the basicknowledge of Retinex theory in detail, as well as some of the classic Retinex models. At the sametime, according to the center surround Retinex model, we propose an improved multi-scale Retinexalgorithm, which mainly to reduce the computation time, and preserving the color of image bydown sampling, mean filtering and a new quantization operation.