KEYWORDS: Light sources, Image compression, Digital photography, Image segmentation, Distortion, Image quality, High dynamic range imaging, Neodymium, Electronic imaging, Current controlled current source
In a backward compatible HDR image/video compression, it is a general approach to reconstruct HDR from compressed
LDR as a prediction to original HDR, which is referred to as inverse tone mapping. Experimental results show that 2-
piecewise 2<sup>nd</sup> order polynomial has the best mapping accuracy than 1 piece high order or 2-piecewise linear, but it is also
the most time-consuming method because to find the optimal pivot point to split LDR range to 2 pieces requires
exhaustive search. In this paper, we propose a fast algorithm that completes optimal 2-piecewise 2<sup>nd</sup> order polynomial
inverse tone mapping in near constant time without quality degradation. We observe that in least square solution, each
entry in the intermediate matrix can be written as the sum of some basic terms, which can be pre-calculated into look-up
tables. Since solving the matrix becomes looking up values in tables, computation time barely differs regardless of the
number of points searched. Hence, we can carry out the most thorough pivot point search to find the optimal pivot that
minimizes MSE in near constant time. Experiment shows that our proposed method achieves the same PSNR
performance while saving 60 times computation time compared to the traditional exhaustive search in 2-piecewise 2<sup>nd</sup>
order polynomial inverse tone mapping with continuous constraint.
In this paper, we propose an adaptive upsampling filter to spatially upscale HDR image based on luminance range of the
HDR picture in each color channel. It first searches for the optimal luminance range values to partition an HDR image to
three different parts: dark, mid-tone and highlight. Then we derive the optimal set of filter coefficients both vertically
and horizontally for each part. When the HDR pixel is within the dark area, we apply one set of filter coefficients to
vertically upsample the pixel. If the HDR pixel falls in mid-tone area, we apply another set of filter for vertical
upsampling. Otherwise the HDR pixel is in highlight area, another set of filter will be applied for vertical upsampling.
Horizontal upsampling will be carried out likewise based on its luminance. The inherent idea to partition HDR image to
different luminance areas is based on the fact that most HDR images are created from multiple exposures. Different
exposures usually demonstrate slight variation in captured signal statistics, such as noise level, subtle misalignment etc.
Hence, to group different regions to three luminance partitions actually helps to eliminate the variation between signals,
and to derive optimal filter for each group with signals of lesser variation is certainly more efficient than for the entire
HDR image. Experimental results show that the proposed adaptive upsampling filter based on luminance ranges outperforms
the optimal upsampling filter around 0.57dB for R channel, 0.44dB for G channel and 0.31dB for B channel.