Research On Image Segmentation And Its Region Of Interest Extraction

In recent years,electronic imaging equipment has been widely used in various industries.With the development of science and technology,the image pixel continues to improve and the pixel size is also increasing,which resulting in the image dataset is too large.The traditional image segmentation algorithm has been unable to meet our demand.The present researchmainly aimed at how to improve the accuracy of segmentation and algorithmefficiency,and often ignore the noise of image.But the image affects result of segmentationin the course of image processing,there are many inherent specialties and unpredictable complexity,the noise in the affects the segmentation effect.Aiming at the above problems,this paper has provided two main works which includes:(1)aiming at theproblems of the traditional segmentation methods,this paper proposes an improved method.Firstly,optimizing the entropy of the target image,whichaim to enhance the image contrast;secondly,using the Wiener Filter compensation to reduce the image noise after enhancing;finally,using Otsu method on region of interest segmentation and extraction,the segmentedimage can not only keep higher information entropy,but also better reflect the details of texture.The results indicate that segmentation accuracy has a certain improvement.(2)Due to camera light uneven orlong exposure time,the gray value of the image is unstable and changes in a larger range,So that the image gray is very dark,the details of the image information can't be displayed.In view of this,the paper provides a novel method.Firstly,the image is grayed;secondly,adoptingHomomorphic Filter to eliminate the influence of illumination;finally,using the fuzzy C-means clustering method for image segmentation.The results show that the information entropy and segmentation error after image segmentation are better than histogram method and K-means clustering method.