CAJ | 학술논문

提出一种基于小波降噪、多尺度图像分割和颜色传递的遥感图像彩色化增强方法。该方法利用一种多尺度图像分割技术，将图像分割为若干子区域，在子区域之间进行亮度匹配和颜色传递，减少了颜色传递误差，再与采用软阈值方法的复数小波包降噪预处理相结合，来增强遥感图像的细节分辨能力。试验结果证明，基于图像分割和色彩传递相结合的方法，其参考彩色图像不需要与目标图像配准，只要特征、风格相近似，就可以对纹理结构相对较简单的全色遥感图像实现全自动的彩色化增强。彩色化的结果图像在美观程度及目标的可识别度上都有了明显增强。
제출일충기우소파강조、다척도도상분할화안색전체적요감도상채색화증강방법。해방법이용일충다척도도상분할기술，장도상분할위약간자구역，재자구역지간진행량도필배화안색전체，감소료안색전체오차，재여채용연역치방법적복수소파포강조예처리상결합，래증강요감도상적세절분변능력。시험결과증명，기우도상분할화색채전체상결합적방법，기삼고채색도상불수요여목표도상배준，지요특정、풍격상근사，취가이대문리결구상대교간단적전색요감도상실현전자동적채색화증강。채색화적결과도상재미관정도급목표적가식별도상도유료명현증강。
This paper presents a newframework of remote sensing images enhancement based on wavelet denoising ,multi‐scale image segmentation and color transfer .A multi‐scale spectral image segmentation algorithm segments images into several sub‐regions .We match luminance and transfer color between the sub‐regions in order to reduce the error .The framework also involves a wavelet denoising preprocess to enhance the ability of distinguishing details of images .The experimental results demonstrate that the method can auto colorize those remote sensing images without registration which have simple textures and structures .The results not only look better in visual but also improve the recognition of objects .