Record some of the process of learning the MIL image processing library.
In Edge Finder source code:
// 分配一个边缘检测缓存.
MIL.MedgeAlloc(MilSystem, MIL.M_CONTOUR, MIL.M_DEFAULT, ref MilEdgeContext);
// 分配一个结果缓存.
MIL.MedgeAllocResult(MilSystem, MIL.M_DEFAULT, ref MilEdgeResult);
// 下面这两句是自定义计算一些边缘特征：M_MOMENT_ELONGATION和M_FERET_MEAN_DIAMETER
MIL.MedgeControl(MilEdgeContext, MIL.M_MOMENT_ELONGATION, MIL.M_ENABLE);
MIL.MedgeControl(MilEdgeContext, MIL.M_FERET_MEAN_DIAMETER + MIL.M_SORT1_DOWN, MIL.M_ENABLE);
// 计算边缘和特征
MIL.MedgeCalculate(MilEdgeContext, MilImage, MIL.M_NULL, MIL.M_NULL, MIL.M_NULL, MilEdgeResult, MIL.M_DEFAULT);
// 返回获取的边缘数量.
MIL.MedgeGetResult(MilEdgeResult, MIL.M_DEFAULT, MIL.M_NUMBER_OF_CHAINS + MIL.M_TYPE_MIL_INT, ref NumEdgeFound);
// 绘制边缘轮廓到图中。
MIL.MgraColor(MIL.M_DEFAULT, EdgeDrawColor);
MIL.MedgeDraw(MIL.M_DEFAULT, MilEdgeResult, GraphicList, MIL.M_DRAW_EDGES, MIL.M_DEFAULT, MIL.M_DEFAULT);The above code only introduces the general steps of edge detection, and there are still many parameters that need to be modified that need to be changed with the actual image. The following is an introduction to the most commonly used MedgeControl in Edge:
MedgeControl Description This function sets the specified control for the edge finder context or edge finder result buffer. For Edge Finder contexts, these settings control the execution of the MedgeCalculate operation and select which edge features MedgeCalculate should calculate. For Edge Finder result buffers, these settings control post manipulation of results. For example, to paint the scaled region of the source image used to compute the result, the paint control values must be set appropriately.
MedgeControl parameters:
MIL.MedgeControl(MilEdgeContext, X, Y);//X, Y represent two parameters
| X | describe | Y | describe |
| M_ACCURACY | Set the precision of edge extraction | M_DISABLE M_HIGH |
DISABLE means disabled and will be calculated with pixel precision. HIGH means to specify very high sub-pixel precision to calculate. |
| M_ANGLE_ACCURACY | Sets the accuracy of corners during edge detection | M_HIGH M_LOW |
Just take it literally |
| M_CHAIN_ALL_NEIGHBORS | How to set up the edge chain | M_DISABLE M_ENABLE |
Controls whether to enable this parameter, which means to use more information to build edges. As shown below. |
| X | describe | Y | describe |
| M_DETAIL_LEVEL | Set the detail level of image extraction, the lower the detail level will find more edges, of course, these edges are not necessarily accurate. | M_HIGH M_MEDIUM M_VERY_HIGH |
Set the level of detail to medium, high, and very high. |
| M_FILTER_SMOOTHNESS | Set the smoothness processing of the image, that is, the degree of denoising. If it is too large, it will change the edge information of the image, and if it is too small, it will not be able to remove noise. This value needs to be set reasonably. (great impact on edge detection) | M_DEFAULT VALUE |
The default value is 50, and the setting range of VALUE is [0, 100] |
| M_FILTER_TYPE | Set the operator for edge detection. The default operator used in MIL is SHEN-CASTAN, which is the heavy operator. | M_DERICHE M_FREI_CHEN M_PREWITT M_SHEN M_SOBEL |
For several common operators, give a formula in the figure below. |
| X | describe | Y | describe |
| M_FLOAT_MODE | Sets whether to force all processing of edges to be computed using floating point precision. | M_DISABLE M_ENABLE |
none |
| M_MAGNITUDE_TYPE | Sets how the magnitude of the edge gradient magnitude is computed at each edge. For CONTOUR, magnitude is the norm of the edge position gradient vector. CREAST is the maximum value of edge positions. | M_NORM M_SQR_NORM |
Default: M_SQR_NORM for contour and M_NORM for CREAST. M_NORM refers to the gradient magnitude, and M_SQR_NORM refers to the square of the gradient magnitude. |
| M_OVERSCAN | Set the scan type for processing boundary images (IIR filter, this parameter will be ignored, such as SHEN and deriche) | M_DISABLE M_MIRROR |
I don't understand this parameter |
There are many more, let’s write here first. Slowly add later~