huff

说明：  一些电路图，包括电脑红外遥控，调速风扇，无线免提电话电路，学习型红外遥控终端，智能脉冲自行车充电器等。(Some circuit, including computer infrared remote control, fan speed, wireless hands-free telephone circuits, a learning IR remote control terminals, smart pulse bicycle chargers.)

： 对一种基于图像几何重心的特征点提取算法进行了改进 改进后的算法可以提取任何边缘线上凹凸处的特征点 其基本 思想是，首先提取图像的边缘轮廓线，计算封闭边缘线和非封闭边缘线的几何重心 然后以重心为极点，对边缘点极坐标化，形成 幅角-极径曲线，在该曲线上寻找局部极大极小值点 最后根据局部非极大（小）抑制原则对极大（小）值点进行取舍获得特征点 模 拟结果表明算法计算简单 稳定性好 该方法提取的特征点能准确地反映目标的形状，对目标的识别 跟踪以及三维表面重建等都 具有重要的价值(This paper modifiesakindoffeatureextractingalgorithmbasedonimageedge.First， imageedgeoutlineisextracted. Second， geometricgravitycenterofedgeoutlineiscalculated.Third， argument-polarradiuscurveofpolarcoordinatesystemwhose poleisgeometricgravityisformed， andmaximumandminimumvaluepointsaresearched.Atlast， featurepointsareobtainedby selectingmaximum （minimum） valuepointsaccordingtolocalnon-maximum （minimum） depressionprinciple.Theresultindicates thatthealgorithmissimpleanditsstabilityisgood.Itisimportantforobjectrecognitionandthree-dimensionreconstruction.)

首先将非平稳的故障振动信号进行双树复小波包分解，得 到不同频带的分量；然后对每个分量求其峭度值和相关系数并进行比较；最后选取峭度值和相关系数较大的分量 进行软阈值降噪和双树复小波包重构，即可有效地消除振动信号中噪声的干扰，同时保留信号中的有效信息即实 现了故障特征信息的提取。(In view of the above situation, a new fault diagnosis method is proposed based on dual-tree complex wavelet packet transform and threshold de-noising. Firstly, the non-stationary fault signal is decomposed into several different frequency band components through dual-tree complex wavelet packet decomposition. Secondly, Kurtosis and the cross-correlation coefficient of each component are obtained and compared. Due to the kurtosis reflecting the signal variations, if the kurtosis value is bigger, the degree of the change of signal is bigger too. The correlation coefficient can reflect the proximity between the component and the original signal at the same time, the correlation coefficient is bigger, the more similar with the original signal. Finally, the components that have a bigger value are chosen to be de-noised by a soft threshold and reconstructed by dual-tree complex wavelet packet transform. The noise interference was eliminated effectively, and the effective si)

VC图象处理电子书，Chm格式，基本与位图相关的都作了说明，很不错(VC image processing e-books, Chm format, and basic bitmap were made related to the note, a very good)

说明：  这篇论文提出了一种较好的水下图像增强的方法。首先使用经过端到端训练的卷积神经网络去测量输入图片，同时以自适应双边滤波器对传输图片进行处理。接着提出一种基于白平衡的策略来消除图片的颜色偏差，用拉普拉斯金字塔融合获得无雾和色彩校正图像的融合结果。 最后，输出图像被转换为混合小波和方向滤波器组（HWD）域，用于去噪和边缘增强。 实验结果表明，该方法可以消除颜色失真，提高水下图像的清晰度。(This paper proposes a better underwater image enhancement method. Firstly, an end-to-end training convolutional neural network is used to measure the input image, and an adaptive bilateral filter is used to process the transmitted image. Then a strategy based on white balance is proposed to eliminate the color deviation of images. The fusion results of fog-free and color correction images are obtained by Laplacian pyramid fusion. Finally, the output image is converted into hybrid wavelet and directional filter bank (HWD) domain for denoising and edge enhancement. The experimental results show that the method can eliminate color distortion and improve the clarity of underwater images.)

KLDIV Kullback-Leibler or Jensen-Shannon divergence between two distributions.

说明：  能实现图像的分割和识别,能够区分开近似颜色区域的影响(this code can perform image crop and recongition)

基于机器视觉，控制三自由度运动控制台来调整显微镜位置，通过图像处理得到易拉罐的表面刻痕大小和长度。(The software control the motion platform with 3 d.o.f. to position the microscope and then test the notch of cans using image processing.)

产生一个验证码，有四个数字组成，有噪音，椒盐噪声等(failed to translate)

说明：  图像边缘检测及图像区域分割、目标检测、目标识别(Image edge detection, image region segmentation, target detection and target recognition)