Retinex-based image enhancement method in a dark vision environment

Abstract

The invention belongs to the field of image enhancement and computer vision, and particularly relates to a Retinex-based image enhancement method in a dark vision environment, which comprises the following steps: acquiring image data, and dividing the image data into a global illumination image and a local illumination image; carrying out weighted average on corresponding pixel points of the two images to obtain a preliminary estimation illumination image, carrying out edge-preserving smooth filtering on the preliminary estimation illumination image by adopting improved weighted guided filtering, and carrying out improved adaptive Gamma correction on the filtered illumination image to obtain a corrected estimation illumination image; calculating reflection images of an R channel, a G channel and a B channel by adopting a Retinex algorithm and utilizing the original image and the corrected estimated illumination image, and synthesizing the reflection images of the three channels to obtain an enhanced image. Bright area details can be reserved and halo effect can be reduced while images in the dark vision environment are enhanced, excessive enhancement of the images is avoided, and the images in the dark vision environment can be conveniently recognized, detected and the like.

Description

technical field [0001] The invention belongs to the field of image enhancement and computer vision, and in particular relates to a Retinex-based image enhancement method in a dark vision environment. Background technique [0002] Image enhancement technology is widely used in various image processing such as aerospace, military reconnaissance, face recognition, and medical diagnosis. However, there are still problems in image enhancement in dark vision environment, and good image enhancement effect cannot be achieved. From the perspective of human vision, when in a very dark environment, when the luminance is less than a few thousandths of candela / square meter, the cone cells lose their activity, and the rod cells function as light-sensing. The vision at this time is called scotopic vision . In the dark vision environment, due to insufficient light in the night environment, human vision cannot distinguish the structure of the image, but the structure of the image exists ph...

AI Technical Summary

Problems solved by technology

Although the Retinex method can obtain better visual effects, there are phenomena of excessive enhancement and color transfer

In the dehazing enhancement algorithm, the image dehazing model is solved through the dark channel prior theory, and the low-light video inversion is brought into the dehazing model to obtain a better image enhancement effect, but the basic model that the algorithm relies on is lacking. physical explanation

Images