基于深度学习的图像描述与增强算法研究

Image Description and Enhancement Algorithm Based on Deep Learning

图像描述是一种图文融合技术,旨在用文字对图像内容进行概括性描述。笔者首先提出融合混合注意力机制的图像描述算法,对图像特征进行提取,随后运用对抗逆强化学习的图像描述模型对图像进行描述,然后在LSTM方法的基础上提出OIPM模型,对图像照射分量进行估计,并通过曝光图像进行修正、补充与融合,最后结合局部结构化融合、色度加权融合机制对带融合图像进行融合,最终获取有效描述与增强的图像。

Image description is a kind of image-text fusion technology,which aims to describe the content of the image in words.Based on image description and image enhancement overview,first of all,the image description algorithm of fusion and mixed attention mechanism is proposed to extract the image features,and then the image description model of adression-resistant reinforcement learning is used to describe the image.Secondly,based on the LSTM method,the OIPM model is proposed to estimate the exposure component of the image,and then the exposure image is corrected,supplemented and fused.Finally,combined with local structured fusion and chromaticity weighted fusion mechanism,the image with fusion is fused to obtain the effective description and enhancement of the image.