Target seed searching method based on dynamic sliding window

Abstract

The invention discloses a target seed searching method based on a dynamic sliding window. The method comprises steps: in a pre-processing stage, a training set target feature histogram is extracted as a seed matching basis; in the case of seed searching, the feature matching similarities of current horizontal large window and vertical small window images are calculated respectively; a vertical small searching window stage is entered, and based on a sliding model, the sliding step of the vertical small searching window is determined; when the small searching window arrives at the lower edge of the image, a horizontal large window searching state is entered; the horizontal large searching window calculates the image feature matching similarity firstly; based on a sliding model, the sliding step of the large window is determined until the large window arrives at the right edge of the image; and when the horizontal searching window and the vertical searching window arrive at the right edge and the lower edge at the same time, searching is ended, and a result is outputted. The problem that a cumulative similarity equation commonly used in the sliding window has an over quick redundancy value growth rate and the no-seed window exclusion ability is limited can be overcome, and the efficiency of using the sliding window for target seed searching is greatly improved.

Description

technical field [0001] The invention relates to the field of target seed search methods, in particular to a dynamic sliding window target seed search method. Background technique [0002] Object segmentation using seed growing method is an important method in image segmentation. Designing an efficient seed search method can greatly improve the speed and accuracy of object segmentation, which is of great significance for the practical application of object segmentation. There are generally two solutions to this problem: one is the manual marking method widely used in interactive segmentation, that is, manually interactively delineate the segmentation range or specify the location of the seed; the other is to automatically search through feature detection. The seed needed to split the target. In order to improve the automation degree of object segmentation, the present invention proposes an object seed sub-search based on a dynamic sliding window. [0003] In many practical...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide a method for searching target seeds based on a dynamic sliding window, so as to solve the problem that the redundant value of the sliding window increases too fast and the ability to exclude the non-seed window is limited

Images