search

LWT3437 / LANet

Source Code for “LANet: A Luminance Attentive Network with Scale Invariance for HDR Image Reconstruction” (PG 2021)
16 stars 3 forks source link

about train dataset #5

Open sunyclj opened 2 years ago

sunyclj commented 2 years ago

Hi,May I ask you a question? in your article，‘HDR images stored in relative luminance are scale-invariant, which means the HDR images will hold the same information when multiplied by any positive real number. Based on this observation, we propose a novel normalization method called " HDR calibration " for HDR images stored in relative luminance, calibrating HDR images into a similar luminance scale according to the LDR images’ ,but in wiki,the definition of HDR is 'Information stored in high-dynamic-range images typically corresponds to the physical values of luminance or radiance that can be observed in the real world.', from my perspective,HDR stores data that should be absolute values.Is my understanding correct?what dataset are you using?

LWT3437 commented 2 years ago

@sunyclj Ideal HDR images should be saved as absolute luminance values (the values have specific physical quantities), but this doesn‘t mean that HDR images can only save absolute luminance information. In fact, in the application of computer graphics, we usually do not pay attention to the physical brightness of the light, but to find a suitable scale to render the virtual scene.

The dataset we use has been presented in the supplementary material after the paper

sunyclj commented 2 years ago

@LWT3437ok, I got it.thanks.