Dynamic Downsampling Image Segmentation Method

Abstract

The present invention provides a dynamic downsampling image segmentation method, which is a fast and complete segmentation method for low-contrast images in a strong noise environment. The method is based on the idea of ​​dynamic downsampling and mainly includes the following steps: defining local two-dimensional entropy; The image is dynamically down-sampled to obtain the optimal down-sampling factor; the original image is down-sampled using the optimal down-sampling factor; the image after the optimal down-sampling is subjected to binary segmentation; the coordinates of the segmented target area are mapped back to the original image, achieve the final target segmentation. The core idea of ​​the above method is the dynamic downsampling process, and the local two-dimensional entropy will be calculated for each downsampling result by successively decreasing Δs based on the size of the original image. In this process, the segmentation effect corresponding to the local two-dimensional entropy peak point not only overcomes the influence of irrelevant interference factors in the image on the image segmentation, but also retains the key details of the image, which is the best image we are looking for. sampling factor.

Description

technical field [0001] The invention relates to a dynamic down-sampling image segmentation method, which is a low-contrast image segmentation method in a strong noise environment, especially the rapid and complete segmentation of in-situ plankton images under unsatisfactory imaging conditions. Background technique [0002] Regions of Interest (ROIs) are image regions marked with rectangular frames, ellipses, or other irregular polygons after selecting image pixels based on certain calculation rules in computer vision and image processing tasks. . The extracted ROIs often contain the target to be analyzed. This method of only selecting the target area in the image can reduce the impact of other irrelevant background components on the further analysis of the target in subsequent steps, such as feature extraction and target recognition. and tracking applications. Traditional ROI extraction methods are directly based on the original Figure II Value segmentation is implemented...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to solve the problem that the segmentation process in the prior art is very time-consuming, and for images with strong noise and low contrast, the target area is broken, and the segmentation effect is not ideal, and a dynamic down-sampling image segmentation method is proposed

Images