A template selection and accelerated matching method for nonlinear color space classification

Abstract

The invention provides a template selection and accelerated matching method for nonlinear color space classification, and the method comprises the processes of the model training and the image matching, wherein the model training comprises the steps of collecting a training image sample, extracting a CIE chromaticity diagram of the training image sample, and carrying out the manual marking of a color category number to which the CIE chromaticity diagram belongs; and obtaining a five-layer feedforward neural network model; and the image matching process comprises the following steps of inputting a pair of color images, and setting a sampling rate; carrying out isolated point downsampling processing; obtaining a classification result set; calculating a metric value of the similarity probability; selecting i corresponding to the first k values with the highest scores as a preferred color category number, and correspondingly obtaining a template area in the template image and an area to bematched in the image to be matched according to the preferred color category number; and matching the template area in the template image with the to-be-matched area in the to-be-matched image. Experimental results show that the method has the higher registration rate and execution speed, and the problem that color space color distance measured by a linear model is inconsistent with human eye vision judgment in an existing matching method is solved.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a template selection and accelerated matching method for nonlinear color space classification. Background technique [0002] Template matching algorithms typically consider all possible transformations, including rotation, scaling, and affine transformations. Alexe et al. provide a computationally efficient way to deal with high-dimensional vectors in the matching windows of two images, which extracts the boundary of the overlapping part of the two windows and uses it to constrain and match multiple windows. Tsai et al. proposed to use wave decomposition and ring projection to improve the matching accuracy, and focused on the rotation transformation. Kim et al. proposed a grayscale template matching method, which has better resistance to rotation and scale transformation. Yao et al. propose a method for searching color textures that also takes into account rotation a...

AI Technical Summary

Problems solved by technology

CFAST-Match was proposed by Jia et al. It improves the accuracy of color image template matching by calculating the proportion of different colors in the template area, but this method needs to

Images