has not yet get this part, only some time to understand the simple opencv, also use it to write a few lines of code play an avi, display a few images, another step by step according to the online tutorial for a small demo face recognition, but the principle of what all is not clear, seems to be very troublesome,
1. Do not know much about the so-called "extraction", the result is what? A picture of a only a thin line, or the eigenvalues of the wire?

Extract meaning is a picture of a only a thin line, fine line parts, other parts of the pixels value is 0

CodePudding user response:

refer to 7th floor chengbar response:

silhouette method, two different or picture?

hello, no two pictures, only a picture under the condition of extraction

CodePudding user response:

reference 5 floor zhao4zhong1 reply:

or use vertical direction along the texture radial blur?

Miss zhao, can you tell me what you have said which algorithm fuzzy corresponding? Opencv has a corresponding method?

CodePudding user response:

Is there any limit the width of the line? Is there a relatively fixed direction? If the scope is very small, and the direction is fixed, feel it is good to do some, otherwise might have some trouble, according to some time on the understanding of the convolution, can adopt the similar steps, the following is my understanding of solution:
1. Set a line width of 2 times square box (estimated 2 times more appropriate, or between 1.5 to 2 times), side length should be at least 3 pixels, set the cross position (width and line width) weights 2, the other is 1, the setting position of aim is to make the lines highlight, and other places more dark
2. Take the window sliding over the image and set for each pixel, slip is used to set the weight of calculating the average of all pixels within the scope of a formed a pixel value, after sliding all pixels, to form a picture of two new image pixel is smaller than the original image
3. The new image binarization processing, calculating the average of all the pixels, and then put all less than average pixel is 0, big is set to 255
4. The new image, put all the image edge and not connect 255 pixels are all set to 0

Haha, I don't know whether to work, is also a method of fuzzy again after processing, just can't fuzzy using simple readme average
If you have time I try to write a program

CodePudding user response:

CodePudding user response:

The

reference 13 floor zhao4zhong1 reply:

miss zhao, how to implement this motion blur

CodePudding user response:

Baidu search "motion blur the source code,"

CodePudding user response:

Attention and study