Combining recursive Retinex and eye adaptation for high dynamic range compression

A novel adaptation-based recursive Retinex algorithm is proposed in this paper. The recursive scheme works in a similar way to the first recursive Retinex algorithm proposed by Frankle and McCann. Simulating the spatial response of retinal photoreceptors, the pixel level adaptation is implemented instead of the ratio-product operations. Thus the dynamic range is compressed efficiently without altering the spatial properties of the original image. In addition, the iteration proceeds in a simultaneous manner, thus the spatial distortion over directions is eliminated. Moreover, the sensitivity constant can be tuned for different strength of local contrast enhancement. Finally, global dynamic range compression and local contrast enhancement are achieved simultaneously.