Research Of Image Enhancement Methods

Image enhancement is a main task in image processing for the purpose of improving visual effect, at the same time, serving for image recognition, image understanding and so on. Image processing is an ill-conditioned problem. At present, the main method to ill-conditioned problem is normalization, namely constrained conditions are added to it. But the strong randomicity of images determines that it is difficult to find the constrained conditions adapting to multi-type images and the whole image. For this reason, several hypotheses are applied in image processing algorithms, but research results indicate that the more hypothesis is in algorithm, the worse the performance is. Adaption of image processing is a reasonable solution, because hypotheses will reduce and are closer to practice aiming at different instances. This thesis proposes an algorithmic adaptive image enhancement method based on pixel classification, but this method only can change edge slope and can not change edge contrast. For this reason, region histogram equalization is advanced in the thesis, this method can improve edge slope and contrast at one time, furthermore, it can restrain noise in smooth regions.Image information measure indicates some basic properties in pixel classification, it could classify the pixels into smooth pixels, edge pixels and texture pixels. The magnitude of image information measure reflects the intensity property of pixel edge. The author make a research on image information measure and has the idea of gaining more information from image information measure, so the definition of image information measure is improved. Abundant image pixel information in the improved image information measure is used for algorithmic adaptive enhancement of form image. The idea of algorithmic adaptive image enhancement is confirmed by the experiment and the image quality is quantificationally evaluated using the measure.The region histogram equalization is different from traditional histogram modification technology, the idea is based on human vision character and edge effect in image vision. The method divides the image into regions by the way of edge connectionbased on image information measure, considers the areas as equalization objects, makes the area display on image gray scale to improve the whole image gray scale scope to enhance edge, meanwhile, in every region, it keeps every pixel's gray level difference relative to the region mean value invariably, which will not introduce noise to every region, so the region's smoothness is preserved. Finally, the author designs a filtering method for the noise produced by the algorithm.The author advances his own realized method concerning the two preceding enhancement ideas and makes a lot of experiments to verify the two ideas. All of the pictures in the thesis are simulation results.