Research Of The Methods Of Removing The Cloud And Fog In Optical Images

With the rapid development of the digital computer and the space technology, the digital image processing technology has made a great achievement, and has become one of the important methods for obtaining more effective information about the environments. In particular, the optical image processing based on the remote imaging technique can be applied to many fields, such as the military, the meteorology, the environment, the forestry, the agriculture, and the geology etc. The optical images by remote imaging system are affected easily by the weather and the influences of the cloud and fog are more universal. As a result of the ground and objects that covered by cloud and fog in large region, therefore we can not obtain the clear information of the ground and the objects. For promoting the efficiency of use of the images, we need to seek an effective method for reducing or removing the influence of the cloud and fog on the image quality.Researchers have proposed many methods for removing the influence of the cloud and fog on the optical images, among which the multi-spectrum image, the multi-image interpolation, multi-sensor data fusion, and Homomorphic filtering methods have met with success. In recent years, the Retinex algorithm has been used for removing or reducing influence of the cloud and fog on the images and is very effective. In this thesis, the methods for enhancing the single optical image affected by the cloud and fog are investigated. The methods for removing the cloud and fog are proposed based on the weighted wavelet coefficients and the wavelet packet decomposition threshold. The experimental results obtained by using the wavelet transform, the homomorphic filtering, and the Retinex algorithm are compared. The improved wavelet transform method proposed in this paper is more effective than other two algorithms.The work can be categorized into the following six main aspects:1. The principles of the image degradation and reconstruction are analyzed. The imaging model of an optical system is investigated, based on which the causes and mechanisms of image degradation are discussed.2. The general principle and the method of the enhancement of the optical images affected by the cloud and fog is described, including the histogram processing method, the homomorphic filtering algorithm, and the Retinex algorithm. The characteristics of each algorithm and the results of the image enhancement with each method are given.3. The wavelet transform theory is studied and an effective method is then proposed for removing cloud and fog using the weighted wavelet coefficients based on the multi-resolution characteristics of the wavelet transform. By transforming the digital images with multilayered wavelet, the wavelet coefficients of each wavelet layer as well as the approximate coefficients are obtained. Then according to the wavelet transform theory, the frequency relations between the information which these coefficients represent are analyzed. According to this and combining frequency characteristics between the cloud and fog and the scenery information, the influences of the number of the demarcated levels, weight, and wavelet function on the test results are considered, an improved method for removing the cloud and fog using the weighted wavelet coefficient is proposed. The experimental results have shown that the method of the weighted wavelet coefficient is more effective for removal and reduction of the cloud and fog than the homomorphism filter and the Retinex algorithms.4. As for the optical images partly covered by the cloud and fog, an improved method is proposed to remove the cloud and fog, which is based on wavelet packet decomposition threshold by analyzing wavelet packet transform. In this method, for the higher level wavelet coefficients, this method can distinguish the cloud and fog areas in the images by brightness threshold and then these areas are removed; For the lower level wavelet coefficients, the lower frequency coefficients can be removed for removing the residual cloud and fog information through the further wavelet packet decomposition.5. We propose the color correction method of an image after removing the cloud and fog by using the wavelet transform. According to the brightness changes in the scenery, the cloud and fog regions before and after the image enhancement, we compare the enhanced image with the original image, and keep a darker part in two images, thus our algorithm can not only keep effectively the effect of removing the cloud and fog, but also reduce the degree of the color distortion.6. The evaluation criterion of the quality of the enhanced images is proposed by using brightness, contrast, entropy, gradient, and fidelity etc. For the images of the targets covered partly by cloud and fog, a novel method is proposed for evaluating the effect for removing the cloud and fog using the product of the contrast and the entropy divided by brightness according to the massive experimental results; For the darker brightness images uniformly covered by cloud and fog, the product of brightness, the contrast, and the entropy is used as the basis for evaluating the effect for removing the cloud and fog.