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
-
摘要: 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。Abstract: To solve the problem of high dynamic range image display on traditional devices, a regionalized 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 employed 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 adjustment. Experimental results show that the model can preserve details, accurately predict, and restore color information.