Research On Watershed Segmentation Algorithm With Multi-Scale Morphology & Adaptive Threshold Marker

To suppress the over-segmentation phenomenon caused by noise and texture detail of classical watershed algorithm,analysis and improved research is done for watershed transform application in segmentation technology,a new watershed segmentation algorithm with multi-scale morphology and adaptive threshold marker is proposed.To guarantee integrality of object edge information effectively,this paper applies watershed algorithm directly to the original gradient image but not the simplified image to achieve image segmentation.Firstly the term multi-scale morphology is proposed in algorithm,and morphological gradient of color image is computed with multi-scale theory,which achieves a balance between image simplication and edge protection commendably.Then low frequency of gradient image is obtained using BLPF-2according to illuminance-reflection model,and morphological closing reconstruction is performed for filtering dark noise.The depth H of morphology extended minimum transform is selected by adaptive threshold technique,and marks reconstructed gradient.Next original color image gradient is modified by morphology imposition marker technology.Finally watershed algorithm is applied to modified gradient and gains ideal segmentation result.Evaluate improved algorithm with qualitative and quantitative methods by designing contrast experiments,and do subjective analysis using indices of GCE,PRI and VOI,which proves that the proposed algorithm in this paper can get more consistent results compared with human segmentation,and has stronger adaptability.