High-precision image information extraction method based on low dynamic range

Abstract

The invention relates to a high-precision image information extraction method based on a low dynamic range, and the method comprises the steps: 1, carrying out feature extraction of an image to acquire three RGB channels of an original image and a V brightness channel in an HSV color space; 2, outputting 48 coefficients in groups by using a full convolutional neural network structure, and adding ashortcut structure on the basis of the 48 coefficients to realize fusion of high-level features and low-level features; finally, outputting the 48 spherical harmonic coefficients in total, dividing the 48 spherical harmonic coefficients into 16 groups, wherein each group comprises three pieces of data which represent components on an R channel, a G channel and a B channel respectively; 3, establishing a spherical harmonic coefficient loss function and a diffuse reflection map loss function, and calculating a mean square error loss function and a diffuse reflection map loss function of 48 spherical harmonic coefficients; and 4, performing feedback constraint on the full convolutional neural network structure by utilizing the mean square error loss function and the diffuse reflection mapping loss function of the 48 spherical harmonic coefficients in the step 3.

Description

technical field [0001] The invention is a method for extracting high-precision image information based on low dynamic range, which belongs to the field of image analysis. Background technique [0002] Restoring the information of the original scene from the picture plays a very important role in many applications, such as augmented reality, film post-production, virtual military exercises, image design, interior design, virtual advertisement, virtual full-length mirror, and entertainment games, etc. These applications all include the operation of superimposing virtual objects and virtual scenes on real scenes to enhance and expand the real world. In order to perfectly integrate virtual and real scenes to increase the realism of virtual objects, it is necessary to ensure that the image information of virtual objects and real scenes are consistent. When the image information of real scenes changes, the image information of virtual objects will also change accordingly. In orde...

AI Technical Summary

Problems solved by technology

Although many studies have applied deep neural networks to restore image information from a single indoor or outdoor picture and achieved good results, how to obtain high-precision, high-quality image information parameters is still a problem worth exploring.

Images