基于NLP与Retinex的视频帧间FIF分析研究

Research on Feature Analysis of Video Inter Frame Fuzzy Image Based on NLP and Retinex

为了更好的对视频帧间模糊图像进行特征分析,提出了基于NPL与Retinex相结合的方法。首先从颜色衰减特性、对比度特性和时频域特性三个方面对视频帧间的模糊图像进行研究,通过图像本身特征对模糊图像进行质量评价。然后利用词干提取、词形还原和停用词转移技术对图像特征进行处理,采用HMM模型,通过非线性激活函数完成图像文本的特征提取。最后利用高斯环绕尺度,对图像的输出结果进行加权求和,约束最小二乘滤波色调H对噪声的敏感度,复原图像平滑度量的最优值。实验结果表明,基于NPL与Retinex的视频帧间模糊图像特征分析方法对模糊图像具有较好的评价效果,而且图像更加细腻、模糊图像的色彩均衡性更好。

In order to better analyze the characteristics of blurred images between video frames,a method based on the combination of NPL and Retinex is proposed.Firstly,the fuzzy image between video frames was studied from three aspects:color attenuation characteristics,contrast characteristics and time-frequency characteristics.The quality of the fuzzy image was evaluated through the characteristics of the image itself.Then the image features were processed by stem extraction,word form restoration and stop word transfer technology.The HMM model was used to complete the feature extraction of image text through a nonlinear activation function.Finally,the Gaussian surround scale was used to weighted sum the output results of the image,constrain the sensitivity of the least-squares filtered hue H to noise,and restore the optimal value of the image smoothing measure.The experimental results show that the video inter-frame fuzzy image feature analysis method based on NPL and Retinex not only has a good evaluation effect on the fuzzy image,but also the image is more delicate and the color balance of the fuzzy image is better.