A Study On Boundaries Detection Algorithm Of The Image Based On Multi-feature

The application of digital image processing is becoming more widespread, related to allareas of life and production. The technology of image boundaries detection as one of thedirection of digital image processing is widely used in image segmentation, motion detection,pattern recognition, and other fields. The image boundaries detection is an important part ofimage processing techniques, improve the technical level of image boundaries detection is ofgreat significance for the application of all aspects. In this paper, the research work isbeginning on the study of various features of the image, focusing on the computation of thegradient of the features and the combination of gradient of various characteristics, aims toenhance the results of the image boundaries detection.The image information is shown by the image features of brightness, color and texturefrom different aspects. Usually the boundaries of objects in the image will be more dramaticchanges in the characteristics, and the gradient of the characteristics can reflect the change ofthe image characteristics, so we can use the salient gradient of the characteristics outline ofthe image boundaries significantly. The various features of image are grouped numericallyfirstly, according to the result of grouping us can statistic out the characteristics histogram of apixel and the pixels around it, and then calculate the gradient of multiple directions with thefeatures histograms. Finally, a significant gradient can be selected to characterize the pointfeature information.It is because of the different characteristics reflect the image information of differentemphasis; the image boundaries detected by the gradient of different features have its owncharacteristics: the gradient of brightness is most sensitive to the changes of light and shade ofimage, while color gradient in contrasting color image detection effect is better, and texturegradient distribution is more balanced in the whole image. In order to obtain the advantages ofgradient of the various features, we studied the regression analysis to combine the gradient ofvarious features. By comparing the experimental results, we use multi-feature-based imagecontour detection algorithm to extract image contours obtained the better results.