Power device infrared chart segmentation method based on adaptive quantization enhancement

The invention provides a power device infrared chart segmentation method based on adaptive quantization enhancement. The method decomposes the regions of an image based on regional characteristics, is combined with the Retinex theory to carry out gamma transformation on the illumination component, and compensates noise points by using prior information on the reflection component to realize adaptive quantization enhancement of an infrared chart. The gray dynamic range is improved. A multi-scale structure preserving smoothing algorithm is improved. A target device is segmented based on adaptive quantization enhancement. The method has the advantage of great anti-noise ability, can segment one or more target devices to remove most interference, and prevents the structure of the target devices from being damaged. 本发明提供种基于自适应量化增强的电力设备红外热图分割方法。该方法基于区域特性对图像各区域进行分解，与Retinex理论相结合，在照度分量上进行伽马变换，在反射分量上利用先验信息对噪声点进行补偿，实现红外热图的自适应量化增强，提高灰度动态范围。并改进了多尺度结构保留型平滑算法，在自适应量化增强的基础上对目标设备进行分割。该方法具有较好的抗噪能力，且可以实现对到多个目标设备进行分割，去除多数干扰，且保证了