Multi-object image partitioning method based on grayscale distribution

Abstract

The invention provides a multi-object image partitioning method based on grayscale distribution, belongs to the field of image processing technology and aims to solve the problem that it is difficultto apply an Otsu method to multi-object image partitioning. According to the technical scheme, the problem of determining the number of classifications in a multi-object image is converted into the problem of solving the number of crests and troughs of a curve chart density function; for this purpose, it is needed to perform statistical analysis on the occurrence number of each grayscale value inthe image and draw a wavy grayscale distribution curve chart; and discrete Gaussian convolution is performed on a grayscale distribution curve, meanwhile, convolution of all-order derivatives of a Gaussian function is performed on the grayscale distribution curve, and then the positions of the crests and the troughs are found on the grayscale distribution curve to automatically determine a partitioning number and multiple grayscale thresholds for image partitioning. The method is convenient to operate, quick in implementation and especially suitable for partitioning of the multi-object image.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a multi-object image segmentation method based on gray scale distribution. Background technique [0002] Image segmentation is an important step in many studies such as object recognition. Image segmentation is mainly the process of dividing an image into several regions with unique properties and extracting valuable objects. Foreground objects or target objects extracted through image segmentation can be used in research fields such as image-based semantic recognition and image search. Among the existing image segmentation algorithms, the threshold-based image segmentation method is simple in calculation, fast in speed, high in operation efficiency, and is the most widely used. [0003] Among threshold-based image segmentation algorithms, the method of maximum variance between classes is the most widely used algorithm. The maximum inter-class variance method divides ...

AI Technical Summary

Problems solved by technology

However, for the problems existing in images with a variety of foreground objects and large differences in grayscale, for example, multiple grayscale thresholds are required to distinguish foreground objects with different grayscales, and algorithms such as the maximum between-class variance method are very difficult to solve. Difficult to automatically judge the number of classifications that need to be determined, the gray levels that need to be divided, and the number of objects that need to be segmented, etc.

Images