Image bit depth extension method and device based on hybrid framework

Abstract

The invention discloses an image bit depth extension method and device based on a hybrid framework. By combining the traditional debanding algorithm and the learning algorithm based on the deep network, the unnatural effect of the flat area of ​​the image can be better removed, and at the same time it is more realistic. Restoring the numerical information of the missing bits; including the extraction of flat areas of images, the bit depth expansion of flat areas based on local adaptive pixel value adjustment, and the bit depth expansion of non-flat areas based on convolutional neural networks. The present invention uses a learning-based method to solve the problem of realistically restoring missing bits by training an effective deep network; at the same time, it uses a simple and robust local adaptive pixel value adjustment method for flat areas to effectively suppress the banding of flat areas. Effect, ringing effect, and unnatural effects such as flat noise to improve the subjective visual quality of flat areas.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to image bit depth extension and enhancement processing technology, in particular to an image bit depth extension method and device based on a hybrid framework, which combines adaptive pixel value adjustment and convolutional neural network based reconstruction technique. Background technique [0002] Image bit-depth expansion (Bit-Depth Expansion) refers to the recovery of high-bit-depth images from low-bit-depth images. The bit depth of an image (Bit-Depth) is the number of binary bits in the value of each pixel in the image. For example, the value range of a pixel in an 8-bits image is 0 to 255 (28). The higher the bit depth of the image, the more delicate the brightness changes can be reflected. The bit depth that the human eye can perceive is between 12 and 14 bits, so we often feel that the image seen through the 8bits display is different from the picture seen by the...

AI Technical Summary

Problems solved by technology

However, the current bit-depth extension methods are mostly traditional non-learning methods, which use fixed strategies to fill in the missing bit information, but the ability to reconstruct the real value of the missing bit is weak, and at the same time, it often removes the banding effect. Causes loss of high-frequency detail in non-flat areas

Existing methods are difficult to restore high-bit-depth images without unnatural effects and high-frequency texture details with higher fidelity from low-bit-depth images

Images