摘要大多数遥感图像由于由不同的传感器获取，或者在不同时间、不同光照下获取，图像的照度不尽相同。为使图像的亮度趋于一致以便对比和信息提取，通常采用的方法是对图像进行光照一致性校正。通过对图像增强定义的阐述，引出直方图均衡化、同态滤波、基于梯度域的方法、中心/环绕Retinex方法和伽马校正等图像增强算法。通过对比各方法的优劣，提出了基于图像拼接的光照一致性校正。之后设计了矩形模板和以一幅图做模板的直方图规定化方法，同时提出了能够定量评价光照一致性的参数。实验结果表明，两种不同模板的直方图规定化方法能够达到光照一致性校正的目的，有良好的视觉效果，且定量评价参数能够对光照一致性程度做出正确评定，与主观感受一致。63863

毕业论文关键词  遥感图像；光照一致性校正；图像增强；直方图规定化；定量评价

毕业设计说明书（论文）外文摘要

Title    Illumination normalization for remote-sensing image

Abstract The majority of remote-sensing images are captured by different sensors, or obtained at different time, or under different levels of light. Those images have different illuminations. In order to uniformize the brightness of images for comparison and data extraction purposes, the most common way is to apply illumination normalization to those images. By demonstrating the definition of image enhancement, this paper introduces image contrast algorithms like histogram equalization, homomorphic filtering, gradient domain, centre/surrounding Retinex and Gamma Correction. Comparing the advantages and disadvantages of the resulting images processed by different methods, this paper further raises a method for illumination normalization based on image mosaic. In the meantime, it also analyzes the drawbacks. A later experiment implementing the method of histogram specification,in which the specification was carried out by using a fixed rectangle template and a template with one of the images, together with a parameter quantifying illumination normalization, demonstrates that histogram specification under both situations could obtain illumination normalization and favourable visual effects. In addition, the quantifying parameter could provide correct assessment to the level of illumination normalization, which is in line with subjective perceptions.

Keywords  remote-sensing image, illumination normalization, image enhancement, histogram specification，quantitative assessment

目   次

1  绪论 1

1.1  遥感图像增强的研究背景及研究意义 1

1.2  遥感图像增强的研究现状 1

1.3  光照一致性问题的研究现状及存在问题 2

1.3.1  研究现状 2

1.3.2  存在的问题 4

1.4  本文的研究内容 4

2  光照校正方法 5

2.1  直方图均衡化 5

2.2  同态滤波 9

2.3  基于梯度域方法的图像增强 13

2.4  单尺度Retinex算法 16

2.5  伽马校正 18

3  基于图像拼接的光照增强 20

3.1  具体方法 21

3.2  实验结果与讨论 21

3.3  影响结果的因素