【摘要】 阴影是由于场景的照明不一致而出现在图像中的自然现象。当物体的透光度不同或由不同光源照射会在图像中产生不同强度的阴影。因此,对具有不同阴影强度的图像阴影去除进行了研究。通过用户交互得到阴影掩模,接下来检测具有围绕阴影边界的可变间隔和长度的样本强度分布,这避免了不均匀边界引起的伪影。然后结合超像素分割算法和FCM_S聚类算法对阴影部分进行区域合并,用于阴影和图像边界处尺度的估计。为了恢复阴影部分光照,采用图像修复算法传播阴影尺度场。最后,得到阴影去除后的图像,并进行色彩校正。为了证明该算法的优越性能,与其他算法进行定量比较,结果有所提升。更多还原

【Abstract】 Shadows are natural phenomena that appear in images due to inconsistent lighting in the scene. When the light transmittance of an object is different or illuminated by different light sources, different intensity shadows will be generated in the image. Therefore, research has been performed on image shadow removal with different shadow intensities. Obtain a shadow mask through user interaction, and then detect the sample intensity distribution with a variable interval and length around the shadow boundary, which avoids artifacts caused by uneven boundaries. Then combine the superpixel segmentation algorithm and the FCM_S clustering algorithm to combine regions of the shadows to estimate the scales at the shadows and image boundaries. In order to restore the shadow part of the light, an image repair algorithm is used to propagate the shadow scale field. Finally, the shadow-removed image is obtained and color corrected. In order to prove the superior performance of this algorithm, a quantitative comparison with other algorithms has been made to improve the results.更多还原