分区非局部均值色貌模型在色调映射中的应用

Image Color Appearance Model Based on Regionalized Non-local Means Filter and Its Application to Tone Mapping

针对高动态范围(high dynamic range,HDR)图像的动态范围与传统显示设备不匹配的问题,给出了一种分区的非局部均值滤波,结合色貌模型对色彩进行预测与重构,并应用于色调映射。分区的非局部均值在计算权重之前对图像亮度进行分区,将图像分为基本层和细节层。通过对基本层进行色适应和色调压缩变换,并进行细节补偿操作,保证了再现图像细节可见,最后对图像色彩进行重构,预测原始色彩信息并校正。实验结果表明,该模型可更好地对细节进行保留,准确地预测并还原色彩信息。

To solve the problem of high dynamic range image display on traditional devices, a reglonai- ized non-local means filter is proposed. Together with image appearance model framework, it is used to reconstruct color information and tone mapping. The original non-local means filter was modified by considering region segmentation as a new factor for the weight function. The new filter was em- ployed to decompose an image into base layer and detail layer. The reconstructed image was obtained by several steps including chromatic adaptation, tone compression, detail compensation and adjust- ment. Experimental results show that the model can preserve details, accurately predict, arid restore color information.