CN-116342889-B - Method for quantizing radiation bit depth residual error of hyperspectral image

CN116342889BCN 116342889 BCN116342889 BCN 116342889BCN-116342889-B

Abstract

The invention provides a hyperspectral image radiation bit depth residual error quantification method which mainly comprises the links of image outlier removal, radiation bit depth residual error quantification coefficient determination, bit depth characteristic image generation, residual error image generation and the like. Image outlier removal, namely importing a hyperspectral image to be processed, detecting outliers and removing. And determining the radiation bit depth residual quantization coefficient, namely calculating the maximum radiation value of the hyperspectral image, determining the radiation bit depth residual quantization level of the hyperspectral image, and further determining the quantization coefficient. And generating bit depth characteristic images and residual images, namely obtaining hyperspectral bit depth characteristic images with different radiation bit depths and the residual images thereof. The method is suitable for quantization of various hyperspectral image radiation bit depth residuals.

Inventors

ZHANG AIWU
WANG JUAN
ZHANG XIZHEN

Assignees

首都师范大学

Dates

Publication Date: 20260512
Application Date: 20230117

Claims (2)

1. A method for quantizing a hyperspectral image radiation bit depth residual error comprises the following steps: detecting abnormal values of all wave bands in the hyperspectral image, and replacing the abnormal values by using a neighborhood mean value to obtain a replaced hyperspectral image; Calculating the maximum radiation value of the replaced hyperspectral image, determining the radiation bit depth residual error quantization level of the hyperspectral image, and finally obtaining the radiation bit depth residual error quantization coefficient; Performing residual quantization on the radiation value of each pixel of the hyperspectral image according to the radiation bit depth residual quantization coefficient obtained in the step (2), and generating bit depth characteristic images and residual images with different radiation bit depths; The step of determining the radiation bit depth residual quantized coefficients in the step (2) is as follows: calculating the maximum radiation value of the replaced hyperspectral image; step (22), determining the interval of the maximum radiation value, and further determining the radiation bit depth residual quantization level; step (23), finally determining the radiation bit depth residual error quantization coefficient; In the step (3), performing residual quantization on the radiation value of each pixel of the hyperspectral image, and generating bit depth feature images and residual images with different radiation bit depths includes: determining an equation for calculating the pixel radiation value of the bit depth characteristic image through the calculated radiation bit depth residual error quantization coefficient, and generating a bit depth characteristic image; And (32) determining an equation for calculating the pixel radiation value of the residual image by the calculated radiation bit depth residual quantization coefficient and the bit depth characteristic image obtained in the step (31), and generating the residual image.
2. The method for quantifying the residual error of the radiation bit depth of the hyperspectral image according to claim 1, wherein the detecting the outlier in each band in the step (1) and replacing the outlier with the neighborhood mean comprises: step (11), taking the row and column as a unit, and performing band screening to obtain pixel point radiation values smaller than or equal to 0; And (12) replacing the pixel radiation values in the wave band selected in the step (11) by using the pixel radiation average value of all the points in the neighborhood N to obtain a replaced hyperspectral image.

Description

Method for quantizing radiation bit depth residual error of hyperspectral image Technical Field The invention relates to a method for quantifying a radiation bit depth residual error of a hyperspectral image, which belongs to the field of remote sensing and is suitable for any hyperspectral image. Background The image classification precision is improved, and weak information is highlighted as a target pursued by people. The radiation bit depth residual quantization constructs bit depth characteristic images and residual images with different radiation bit depths by dividing high-resolution radiation information of the hyperspectral image, and deep mining of the radiation information is performed, so that the classification precision of ground features is improved. Although few people are researching the influence of the radiation resolution of the hyperspectral image on the classification of the ground objects, the research only analyzes the influence of different radiation resolutions on the classification of the ground objects from the surface, does not fully excavate the radiation information of the high resolution and does not carry out the residual quantification of the radiation bit depth of the hyperspectral image, so the prior art has the defects and shortcomings that the radiation information of the hyperspectral image is not fully utilized and the influence of different radiation bit depths on the classification of the ground objects is not fully excavated. Disclosure of Invention The technical problems to be solved by the invention are to overcome the defects of the prior art, provide a hyperspectral image radiation bit depth residual error quantization method, determine the radiation bit depth quantization level according to the image radiation resolution, and construct bit depth characteristic images and residual error images with different radiation bit depths, wherein the bit depth characteristic images reduce data redundancy on the premise of ensuring classification accuracy, and the residual error images increase image details under the condition that the classification accuracy is not reduced, highlight the characteristics of certain ground features, and are suitable for quantization of radiation bit depth residual errors of various hyperspectral images. The technical scheme adopted by the invention is that the method for quantifying the radiation bit depth residual error of the hyperspectral image comprises the following steps: detecting abnormal values of all wave bands in the hyperspectral image, and replacing the abnormal values by using a neighborhood mean value to obtain a replaced hyperspectral image; Calculating the maximum radiation value of the replaced hyperspectral image, determining the radiation bit depth residual error quantization level of the hyperspectral image, and finally obtaining the radiation bit depth residual error quantization coefficient; and (3) carrying out residual quantization on the radiation value of each pixel of the hyperspectral image according to the radiation bit depth residual quantization coefficient obtained in the step (2) to generate bit depth characteristic images and residual images with different radiation bit depths. Further, the detecting the abnormal value of each band in the step (1), and replacing the abnormal value with the neighborhood mean value includes: step (11), taking the row and column as a unit, and performing band screening to obtain pixel point radiation values smaller than or equal to 0; And (12) replacing the pixel radiation values in the wave band selected in the step (11) by using the pixel radiation average value of all the points in the neighborhood N to obtain a replaced hyperspectral image. Further, the step of determining the bit depth residual quantization coefficient in the step (2) is as follows: calculating the maximum radiation value of the replaced hyperspectral image; step (22), determining the interval of the maximum radiation value, and further determining the radiation bit depth residual quantization level; And (23) finally determining the radiation bit depth residual quantized coefficients. Further, in the step (3), performing residual quantization on the radiation value of each pixel of the hyperspectral image, and generating a bit depth feature image and a residual image with different radiation bit depths includes: Determining an equation for calculating the pixel radiation value of the bit depth characteristic image through the calculated radiation bit depth residual error quantization coefficient, and generating a radiation bit depth characteristic image; And (32) determining an equation for calculating the pixel radiation value of the residual image by the calculated radiation bit depth residual quantization coefficient and the bit depth characteristic image obtained in the step (31), and generating the residual image. Compared with the prior art, the invention has the advantages that: (1) In the aspe