基于分辨率和动态特征融合的视频图像编码方法  

A Video Image Encoding Method Based on Resolution and Dynamic Feature Fusion

作　　者：温卓方 WEN Zhuofang(Shaanxi institute of technology,College of Art and Design,Xi’an 710300,China)

机构地区：[1]陕西国防工业职业技术学院艺术与设计学院,西安710300

出　　处：《自动化与仪器仪表》2025年第2期28-32,36,共6页Automation & Instrumentation

基　　金：陕西省科技厅自然科学基础研究计划项目(2019JM-3416)。

摘　　要：当前的视频编码技术,只关注对视频的简单压缩和传输,忽略了图像的分辨率和动态特征的重要性,无法根据不同区域的视觉差异进行有针对性的调整,这会导致编码后的视频质量不佳,分辨率降低,难以充分展现视频的动态特征等问题。为此,提出了基于分辨率和动态特征融合的视频智能编码方法。利用帧间差值法对连续的视频图像帧进行差分操作,通过计算不同帧间的像素差异,可以有效识别视频中的动态变化,精确提取视频图像中的动态目标特征。利用改进的卷积神经网络,结合空间信息与通道信息识别和提取视频中的分辨率特征。将提取到的分辨率特征和动态特征进行融合,进一步提升纹理特征的表示效果。在编码树单元中,根据感兴趣区域的纹理特征感知权重,对不同区域的量化参数值进行有针对性的调整。对于重要的区域,分配更高的量化参数值,以保留更多的图像细节和特征信息。对于不重要的区域,分配较低的量化参数值,以实现更高效的压缩。通过实验证明:所提方法在提高图像质量、减小编码误差以及减少编码时间方面表现出了显著的优势,且编码后的清晰度较高。The current video coding technology only focuses on simple compression and transmission of videos,ignoring the im-portance of image resolution and dynamic features,and cannot make targeted adjustments based on visual differences in different re-gions.This can lead to poor video quality after encoding,low resolution,and difficulty in fully displaying the dynamic features of vid-eos.Therefore,a video intelligent encoding method based on resolution and dynamic feature fusion is proposed.By using the inter frame difference method to perform differential operations on consecutive video image frames,the dynamic changes in the video can be effectively identified and the dynamic target features in the video image can be accurately extracted by calculating the pixel differences between different frames.Using an improved convolutional neural network,combining spatial and channel information to recognize and extract resolution features in videos.Integrate the extracted resolution features with dynamic features to further enhance the repre-sentation of texture features.In the encoding tree unit,targeted adjustments are made to the quantization parameter values of different regions based on the texture feature perception weights of the regions of interest.For important areas,allocate higher quantization pa-rameter values to preserve more image details and feature information.For unimportant areas,allocate lower quantization parameter values to achieve more efficient compression.Experimental results have shown that the proposed method has significant advantages in improving image quality,reducing encoding errors,and reducing encoding time,and the clarity after encoding is relatively high.

关 键 词：视频编码 帧间差值 深度特征学习 动态特征 分辨率特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]