Target scale selection method for image multi-level segmentation based on depth seeds

Abstract

The invention discloses a target scale selection method for image multi-level segmentation based on depth seeds. The target scale selection method comprises the following steps: segmenting an input image into a plurality of levels by utilizing a multi-level image segmentation algorithm; constructing a multi-level segmentation tree; representing image features by using a color histogram, a texturehistogram and a regional geometric dimension, and evaluating the quality of the segmented region to obtain a segmentation quality score; finding an optimal segmentation tree; performing depth seed positioning on the foreground class and the background class of the input image; establishing a graph by utilizing an image segmentation region in the optimal segmentation tree, and obtaining a segmentation result through solving the minimum segmentation of the graph. According to the method, foreground seeds are positioned through an improved VGG-19 network, background seeds are positioned through asaliency detection algorithm, and multi-level segmentation of a depth seed processing image is generated; a graph model is designed, and optimal scale selection of an image target is realized by utilizing semantic information provided by a deep learning model based on a multi-scale segmentation region contained in a multi-level segmentation result.

Description

technical field [0001] The invention relates to the technical field of image segmentation, in particular to a target scale selection method for multi-level image segmentation based on depth seeds. Background technique [0002] Image segmentation is an important technology in computer vision. Its purpose is to divide the image into several segments according to the characteristics of the image such as color consistency or texture similarity. Although comprehensive segmentation algorithms have also been proposed in the literature, it remains an open challenge to efficiently segment out meaningful parts of objects to visualize human perception. Everyone has a different definition of "meaningful", and different people will have different ideas about how to correctly segment an image. Many people prefer to segment images into several segments, while others tend to only recognize a few segments of the image content. In this case, a single segmentation result can only be generate...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the insufficiency of the target scale selection of the existing hierarchical image segmentation, the present invention provides a method for selecting the target scale of the image multi-level segmentation based on the depth seed, which overcomes the difficulty in finding the correct target scale for the existing multi-level image segmentation algorithm at a single layer level. The disadvantage of the target segmentation scale, by generating the depth seeds of the foreground class and the background class, the optimal segmentation result is obtained

Images