基于多尺度卷积MSRCR的蒙古族家具纹样增强研究

Research on Mongolian Furniture Pattern Enhancement Based on Multi-scale Convolution MSRCR

针对蒙古族家具纹样采集过程中,受到环境、设备、人为干扰以及纹样自身风化等因素影响,产生对比度低、亮度异常、纹样模糊、色彩失真退化等情况,本文提出一种基于多尺度卷积MSRCR增强的算法。通过二次导向滤波、多通道多尺度卷积、线性加权融合、白平衡等步骤,对MSRCR算法进行改进。结果表明:该方法能有效提升纹样图像质量,改善纹样图像的对比度、亮度,在相关评价指标中也更优。

In the process of collection Mongolian furniture patterns,due to the influence of environment,equipment,human interference and pattern weathering,there are low contrast,abnormal brightness,blurred patterns and color distortion degradation.In this paper,an enhancement algorithm based on multi-scale convolution MSRCR was proposed.The MSRCR algorithm was improved by steps such as secondary guided filtering,multi-channel multi-scale con-volution,linear weighted fusion,and white balance.The results showed that this method can effectively improve the quality of pattern image,improve the contrast and brightness of the pattern image,and obtain better values in the relevant evaluation indicators.