To improve the visibility of fog-degraded images, a method for fog-degraded image clearness based on fuzzy logic was presented. The degraded image was mapped from the gray area to the fuzzy area，and then the fuzzy entropy was calculated to obtain the change rate of a pixel. By choosing the appropriate threshold for the fuzzy entropy，the image pixels were put into two categories，and then achieve different fuzzy feature enhancement operations. Experiment results show that this method can efficiently improve the degradation of image and enhance the clearness of image.

 

 

A kind of method with using of vertebra medical image to deal with image key data was presented. First, used of the method of piecewise linear transform to stretch gray level of image for enhancing contrast ratio between target area and background area or reducing the detail in the target area. Second, used of information of pixel gray level to set up weighing chart, choose target seed pixels and background seed pixels, and then make use of rand-walk algorithm to divide up image so that receive edge of contour line. At last, unit area pixel average gray-level in the range of edge contour line was computed. Through comparing gray value of unit area in the region with bone density which can be gotten from physic testing, the bone density may be received in the checked region.

 

 

Aiming at detecting chromatic moving object from video sequences with fixed scene．HSV color space was adopted to represent each raw frame in video sequences，then it established background frame based on self-adaptive background model by extracting V value in raw frame, and detected moving object based on background subtraction．Motion region binarization incorporated H value in raw frame to filter motion region，in order to obtain the foreground frame in which the special moving object was selected．The experimental results show that the algorithm can achieve the special color moving object detection in the video with fixed scene and several moving objects very well．

Three-dimensional clouds for rendering natural scenes are essential. 3D Simplex noise technology is a new kind of noise generation method which is improved by 3D Perlin noise. 3D Simplex noise technology was used to generate 3D noise gray-scale images. Through the noisy image sharpening and color conversion, the 3D Simplex noise images were converted to blue sky and white clouds images, and finally with time function of the GLSL shading language, the fluttering clouds were achieved. The experiment results show that the 3D Simplex noise is low computational complexity to generate realistic images of the cloud and the generated realistic images of the cloud are effective.

In the SAR image processing, wavelet transform has its unique advantages which analyzes image subtly and has fast algorithm that can resolve and reconstruct the image in short time. Partial differential equation's nonlinear diffusion could preserve the edges and textures, but the large SAR image statistics increase the iterative calculation and decrease the efficiency. The method of combining PDE(partial differential equation)'s nonlinear diffusion and wavelet transformation was proposed to remove speckles of SAR image. The experimental results show that the method can not only work quickly but also keep the image's edges and detail information well.

An new image edge detection method based on the wavelet and the fuzzy logic was presented. Basing the characteristic on high frequency part and low frequency part ,detailed information and averaged information were detected by dyadic wavelet transform and fuzzy logic. The detection results are fused by logic operation. Experiment results show that it can solve the contradiction between the availability and accurate position of edge detection. The edge of image has better complementarity and effect of edge detection is evidently better than classical Canny operator and Mallat operator.

Based on the human visual system, a double image watermarking algorithm in wavelet domain was proposed. A piece of meaningful binary image was taken as watermark. On the basis of the original image analysis, the watermark was appended to appropriately selected parts. And double embedment of the watermark was performed through modifying low-frequency and high-frequency coefficients of the selected parts. It is shown that the image embedded watermark has better transparency, and two watermarks robustness complements each other. The algorithm is more robust for handling general images with JPEG compression and noise images.

Genetic algorithm neural networks is a method that uses genetic algorithms to optimize connection right value to replace gradient decrease method. Joining the simulation annealing algorithm in the evolution process of the genetic algorithm, it has the global optimum and local search capability. At the same time, it can not only enhance the rate and efficiency of algorithmic constringency, but also effectively avoid appearing precocity and plunging into local optimum. The method can completely satisfy the accuracy and speed’s requirements of the real-time system. It was studied that genetic algorithm neural networks joining simulation annealing arithmetic operators after reproduction, crossover or mutation, and was compared their performance of optimize in the genetic neural networks.