The Research And Application Of Color Adaptive Mathematical Morphology

Morphology is used in all aspects of image processing,including edge extraction,feature detection,image filtering,image enhancement,region segmentation,and image restoration because of simple efficient thought and rigorous theoretical basis.The selection of structure element is the key to mathematical morphology.In classic morphology,the size and shape of the structure element are fixed,which often leads to poor results.Therefore,the adaptive mathematical morphology has become a hot spot in morphology research.With the development of communication tools,color images containing more information have gradually replaced two valued and grayscale images in daily life.However,the pixel of color image is a vector.It is difficult to extend the mathematical morphology to color image processing.A new method for vector sorting called difference sorting is proposed which can adjust the main color component automatically.Color adaptive automatic threshold mathematical morphology is presented in this paper.An automatic threshold selection rule is established based on the definition of normalized weight and distance parameter.A new color ordering relation named difference ordering is proposed.An evaluation method by defining the image detail loss rate is proposed to measure the image detail preservation after the morphological procession,which makes up the blank of evaluation index for morphological operators.The definition of normalized weight and distance parameter take both local features of the image and distance between pixels into the consideration of automatic threshold.So the structure element can adjust its shape and size adaptively under the condition of automatic threshold.The difference sorting for vector sorting is proposed,which can adjust main color component and do not produce new color.The influence made by different choice of main color component is avoided.Then the basic color adaptive automatic threshold morphological operators including erosion dilation opening and closing are defined.The results of the proposed operators classic morphology and the existing basic color adaptive morphological operators are compared.The results show the proposed operators perform better in preserving the details of the image and the color fluctuates less.In order to evaluate the proposed operators and other morphological operators objectively,a blind image quality evaluation system is introduced to evaluate the overall quality of the processed image.Then the concept of detail loss rate is proposed to evaluate the local quality of the processed image.The superiority of the automatic threshold morphological operator is verified from both overall quality and local quality.The proposed operators are applied to edge extraction for color images.The results show that no matter inner edge,outer edge or gradient edge extraction,the proposed operators can extract more edges,especially the image detail.