Contour Detection Based on Primaty Visual Model
|Course||Applied Computer Technology|
|Keywords||Contour Extraction Complex texture background HMAX model Non - classical receptive field Suppression|
The contour extraction is a feature extraction method in image processing . The outline contains the shape and structure information , but also has strong robustness , contour extraction is indispensable in target recognition feature extraction method . Contour extraction with a classic edge extraction algorithm compared with good treatment effect , contour extraction algorithm based on the non-classical receptive field inhibition , especially suitable for the extraction of object contours in the complex texture background . The algorithm uses the non-classical receptive field inhibition properties can weaken the edge in the background , thereby enhancing the effect of contour extraction . However , this algorithm is based on two layers of visual model the correct contour extraction is not ideal . HMAX model with the basic structure of five floors of the visual cortex function , but it is mainly used for target identification . For an input image , after processing , the final output is a vector indicating the image feature . Therefore , HMAX model is not suitable for the contour extraction. This article is aware of the strengths and weaknesses based on nonclassical receptive field inhibition contour extraction algorithm HMAX model , a combination of both , Finally, the proposed algorithm and write code . The proposed algorithm , containing all kinds of complex texture background image contour extraction and Canny operator and anisotropic suppression algorithm to compare . Best results through contour extraction plans , assessment parameters table and box plot of the three methods were evaluated , results show that the proposed algorithm improves the accuracy of contour extraction , reducing the error extract the background edge is a good outline extraction algorithm , also laid a good foundation for target recognition .