Method and device for image segmentation

Abstract

An image segmentation method comprises: acquiring a target raw image to be segmented; extracting a geometric characteristic of an area to be segmented in the target raw image; acquiring a color characteristic of each of pixels in the target raw image; performing, according to the extracted geometric characteristic and the color characteristic of each of the pixels in the target raw image, clustering on each of the pixels in the target raw image; and segmenting, according to a clustering result, the target raw image.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image segmentation method and device. Background technique [0002] With the development of image technology, how to segment images becomes more and more important. Traditionally, there are two methods for image region segmentation, one is the segmentation method based on color features, and the other is the segmentation method based on edge features and energy functionals. Among them, the segmentation method based on color features has higher requirements on the depth of color, and if the color features of the segmentation target are not obvious, the segmentation will fail. For example, when the color of the lips is very light and almost consistent with the skin color of the face, the lip area cannot be successfully segmented by this segmentation method. The essence of the energy functional is to first establish a parameter expression describing the characteristics of the re...

AI Technical Summary

Problems solved by technology

Among them, the segmentation method based on color features has higher requirements on the depth of color, and if the color features of the segmentation target are not obvious, the segmentation will fail.

For example, when the color of the lips is very light and almost the same as the skin color of the face, the lip area cannot be successfully segmented by this segmentation method

The essence of the energy functional is to first establish a parameter expression describing the characteristics of the region set, and change the shape of the region by adjusting the parameters. When the defined energy functional takes the minimum value, the geometric figure represented by the expression is completely in line with the actual region edge , its disadvantage is that it is difficult to resume parameter expressions when the geometric characteristics of the region are relatively complex, and there are many parameters, and the large number of parameters itself will make it very difficult to find the minimum value of the energy function, and the operating efficiency of the algorithm will also become very high. Low, and the curves it depicts are generally continuous and smooth, but in fact the boundaries of some segmentation objects themselves are not smooth, so the area segmentation effect is not good

Images