1

Topic: The picture analysis

There are 2 here such pictures On the second the accurate "strip" structure which is not present on the first is tracked. The question - what algorithm can be applied to find these bands?

2

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> the Question - what algorithm can be applied to find these bands? Conversion of Hafa: https://en.wikipedia.org/wiki/Hough_transform

3

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> the Question - what algorithm can be applied to find these bands? The most obvious - 2 Fourier: It is made by program Gwyddion

4

Re: The picture analysis

Hello, Qbit86, you wrote: Q> Preobrazovanie Hafa: https://en.wikipedia.org/wiki/Hough_transform I know About it, but how it to apply in this case? Is not present there either straight lines, or lines, there is only a certain grouping of points.

5

Re: The picture analysis

Hello, Mazay, you wrote: M> the Most obvious - 2 Fourier: It is possible hardly more in detail what exactly means? I meanwhile did not catch communication between two your pictures and that is necessary to me.

6

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> Hello, Qbit86, you wrote: Q>> Preobrazovanie Hafa: https://en.wikipedia.org/wiki/Hough_transform PD> I know About it, but how it to apply in this case? Is not present there either straight lines, or lines, there is only a certain grouping of points. To mark sections with high density of points (it is possible through an actuating threshold) and there will be bands. And then to apply Conversion of Hafa.

7

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> Hello, Mazay, you wrote: M>> the Most obvious - 2 Fourier: PD> It is possible hardly more in detail what exactly means? I meanwhile did not catch communication between two your pictures and that is necessary to me. Idea from diffraction methods (LEED, XRD) - crystals have peaks on spectra. Here too there is a peak on Fourier an image if it in polar coordinates to look. But in this case it is better than autocorrelation to count on space. There generally all will be obvious.

8

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> I know About it, but how it to apply in this case? In a forehead. No, truth. PD> is not present there either straight lines, or lines, there is only a certain grouping of points. It is type of algorithm of voting. You enumerate all straight lines, sorting out with any steps of two parameters: distance on a normal from center, and tilt angles. That is two-dimensional space (it is possible to visualize too as a picture). To each cell of this space you appropriate value: how many bright points of the source image lie almost on this straight line (on visualization  a cell-pixel appropriate brightness). Eventually, if on an initial picture there are explicit bands in resultant "picture" there will be bright points (histogram peaks).

9

Re: The picture analysis

Hello, Mazay, you wrote: M> But in this case it is better than autocorrelation to count on space. There generally all will be obvious. Here: For 3_skel For 24_skel

10

Re: The picture analysis

Hello, Mazay, you wrote: M>> But in this case it is better than autocorrelation to count on space. There generally all will be obvious. M> here: , clearly that they can be distinguished thus, but nevertheless other is necessary to me. It is necessary "almost direct" to select these sections

11

Re: The picture analysis

Here that would be desirable to receive. Not mandatory in the form of straight lines, but here these areas - coordinates (marked not all)

12

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> Here that it would be desirable to receive. Not mandatory in the form of straight lines, but here these areas - coordinates (marked not all) PD> Image: 24_lines.jpg Here I received such lines conversion  walking under the image with step 6464 and a window 128128. Color designated weight - the total of pixels along a line. Further it is possible to filter segments by your criterion. Goes?

13

Re: The picture analysis

Hello, Senyai, you wrote: S> Here I received such lines conversion  walking under the image with step 6464 and a window 128128. Color designated weight - the total of pixels along a line. Thanks, very much even are quite good. Explain that means "with step and a window". I so understand, what conversion do not on all picture, and on a window then you shift it on a step? That is here it http://wiki.technicalvision.ru/index.ph … 0%BD%D0%B5

14

Re: The picture analysis

Hello, Pavel Dvorkin, you wrote: PD> the Question - what algorithm can be applied to find these bands? I will support with a sentence to use conversion of Hafa. Here: it is result of call HoughLinesP from OpenCV (this example).