Study On Several Basic Issues Of Color Image Processing

Image is the main ways and means for people to obtain,transfer and express info.With the rapid progress of computer and electronic technology in the last twenty years,color image input/output equipment of every sorts was developed quickly.Color image had been the mainstream of image,even the multimedia information systems and color image database is widely used.Color image processing technique is demanded by every walk of life and becoming a pressing research area,but there are still some troubles exist.In this dissertation,a study focused on of several basic issues of color image processing technique,such as the color difference measure in RGB color space,the texture features analysis of color image,the self-adaptable edge detection algorithm of color image is carried out and some new algorithms are presented.The main achievements and novelties of this dissertation include:1.Proposing a novel color difference formula based on the vector-angular distance in RGB color space.Based on the study of RGB color space and the compares and analysis of several color difference formulas in RGB color space,this dissertation summarizes three rules of color difference in RGB color space and proposes the conception of the importance of the color component.A new color difference formula in RGB color space is achieved by dynamically adjusting the vector distance and angular distance between a pair of RGB colors.The dynamic coefficient is weighted by the importance of color component.Color quantization experiment of highly complex color images demonstrates that this new color difference formula's performance is better than these widely used color difference formulas in RGB color space.2.Extend the grey level co-occurrence matrix to color co-occurrence matrix, improved the texture features description of color image.In this dissertation,a new conception --- color co-occurrence matrix which based on the theory of grey level co-occurrence matrix was proposed.From the color co-occurrence matrix we can extracted both color and texture features.The experiments results show that the color co-occurrence matrix better than the grey level co-occurrence matrix,fits closely with human perception.3.Improve the capability of LOG operators in edge detection.In order to avoid the defects of LOG operator,By combining the first derivative's maximum and the second derivative's zero-crossing of image gray-scale,we successed in eliminating most of the noise in image and locating the edge accurately. 4.Proposing a method to eliminate the discrepancy between color's component express and holistic express.When the first derivative of the pixel's saturation is too big,we use the saturation of these pixels who adjacent and similar with the pixel to smooth the pixel's saturation.By this means the false edge caused by the discrepancy between color's component express and holistic express is eliminated.5.Proposing a self-adaptable method to detect edge in both grey images and color images.This dissertation remove shot noises in image by self-adaptable smoothness And by using variance of gray-scale to auto-adjust Gassian space coefficient and edge detection threshold,the both grey images and color images's edge can be automatically detectedThe efficient of these method is proved by practices.