Parallel fractural network evolution image segmentation method

Abstract

The invention relates to a parallel fractural network evolution image segmentation method. The parallel fractural network evolution image segmentation method includes steps of 1, decomposing a final scale expected by a user into a plurality of scale sequences from small to large and carrying out follow-up operation for each scale unit in a serialized manner so that a previous small scale unit is outputted as input of a corresponding later large scale unit; 2, dividing an image to be segmented (or an object set) into multiple levels of grids according to a spatial range; 3, executing a corresponding subtask of each grid node; and 4, merging effective segmentation results generated in execution processes after all the subtasks are executed, and using the merged segmentation results as inputs of corresponding larger scale units in the scale sequences until the scale specified by the user is realized. Operation of the steps 2, 3 and 4 is carried out continuously if the scale specified by the user is not realized. The parallel fractural network evolution image segmentation method has the advantages that resource consumption is relieved and parallelized owing to decomposition of the multiple levels of grids and a merging strategy, a processing mode in a traditional single-machine environment with low resource effective-segmentation efficiency is changed, the problem of high occupancy of resources is effectively solved, and parallel segmentation efficiency is improved.

Description

technical field [0001] The invention relates to a fractal network evolution image segmentation algorithm, in particular to a parallel fractal network evolution image segmentation method. Background technique [0002] As an important image processing technology, image image segmentation has been widely used in different fields, and thousands of segmentation methods have been proposed, such as watershed segmentation algorithm, mean shift segmentation algorithm and multi-resolution image segmentation algorithm of Definiens company. Among them, the Fractal Net Evolution Approach (FNEA) algorithm is an effective multi-scale image segmentation method, and it has been proved that it is more effective than other commonly used remote sensing segmentation algorithms that only use pixel spectral information. Good segmentation effect. [0003] At present, parallel decomposition algorithms are mainly divided into domain decomposition, functional decomposition, pipeline, divide and conqu...

AI Technical Summary

Problems solved by technology

For problems such as radar filtering, polarization decomposition, pixel-level classification, and co-occurrence matrix feature calculation, since the algorithm processing unit is a small-scale neighborhood of pixels, and the processing units are independent of each other, the decomposition and merging strategy is relatively simple and can be designed. As for the FNEA algorithm, its processing unit involves pixels in a large image range, and the specific size is not fixed, so the decomposition and merging are more complicated, and the general solution cannot be adopted, and it needs to be designed according to the characteristics of the algorithm

Images