Color image super-pixel segmentation method based on similarity between pixels

Abstract

The invention discloses a color image superpixel segmentation method based on similarity between pixels. The method comprises the following steps: firstly, an image to be segmented is initially clustered; then, a seed point is determined based on the initial clustering; whether a seed point needs to be added is judged; if so, the seed point is added; if so, the seed point is added; an initial superpixel is generated according to the initial clustering and the marker of the seed point; For unmarked pixel points, the defined energy function is used to calculate the energy of the seed point andthe pixels in its searching range, The superpixel marker of the seed point with the least energy is selected as the superpixel marker of the unmarked pixel. Finally, the isolated pixel satisfying thethreshold condition and the isolated very small superpixel are merged into the neighboring superpixels which are most similar to the superpixel marker until the number of the current superpixels reaches the desired number, so as to realize the superpixel segmentation of the image. The method of the invention can obtain the ideal image segmentation effect, and provides higher regularity in the image flat area, thereby improving the quality and the segmentation effect of the image segmentation.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a color image superpixel segmentation method based on similarity between pixels. Background technique [0002] Image is an important source for human beings to obtain information from the objective world and an important medium for transmitting information. With the development of technologies such as computers and the Internet, digital image processing technology plays an increasingly important role in industries such as industry, medical care, and transportation. [0003] Superpixel segmentation is to generate atomic regions of pixels ("superpixels") on the image to be segmented. Unlike traditional rigid pixel representations of images, superpixels provide visually meaningful entities that can be used for image processing and computer Atomic units of vision tasks. Representing images using superpixels instead of rigid pixels can make subsequent processing less computa...

AI Technical Summary

Problems solved by technology

However, the SLIC algorithm only utilizes color and spatial features, so the performance of segmentation will decrease when the color features are insufficient

And SLIC focuses on generating regular superpixels, and boundary regularization and boundary coincidence are often contradictory. The more regular the superpixels are, the harder it is for the superpixel boundaries to fit with the image boundaries; and the generated superpixel boundaries and The higher the image boundary fit, the harder it is for superpixels to ensure regularity

Therefore, when SLIC performs image segmentation, some edge fits are not high, weak boundary areas are not completely divided, and the effect of partial boundary images is not very good, and satisfactory segmentation results cannot be obtained.

Images