1 | Houghlines in java.
|
---|
2 | Houghlines does detect white lines. Use binary_inv to show with houghlineTest.png
|
---|
3 |
|
---|
4 | OTSU threshold gives best black and white response
|
---|
5 |
|
---|
6 | Would it be useful to process what lines have been detected?
|
---|
7 | As in, filter out every line detected except for horizontal lines?
|
---|
8 |
|
---|
9 | BINARYINV seems to be result in a better looking image than canny for sheet music
|
---|
10 |
|
---|
11 | In houghlines parameters, having resoluation of 1 degree and 1 pixel makes sense. Would allow for sheet music image to be at any orientation.
|
---|
12 | Threshold would ideally need to vary dependant on how much sheet music there is.
|
---|
13 | Which brings to the table filtering. Having a good filter that eliminates noise leaves the sheet music free would be good. The binary inv would be suitable, since houghlines detect white lines.
|
---|
14 | Need filter to denoize image. OTSU seems to do a good job.
|
---|
15 |
|
---|
16 | The threshold would need to be adjusted dependant on how many white pixels there are.
|
---|
17 | houghlinesP has start and end points. where houghlines does not. Which is why houghlinesP performs better on a set of images becuz of the end points rather than just "a line must meet a certain threshold at that coordinate for it to be registered"
|
---|
18 | Houghlines in order to be effetive must be scaled to suit the image.
|
---|
19 |
|
---|
20 | HoughlinesP contains this:
|
---|
21 | minLinLength: The minimum number of points that can form a line. Lines with less than this number of points are disregarded.
|
---|
22 | maxLineGap: The maximum gap between two points to be considered in the same line.
|
---|
23 |
|
---|
24 | ____ _____ ______ ______
|
---|
25 |
|
---|
26 | HoughlinesP can calculate on a local scale, comaparing the distance between just the gap in the line.
|
---|
27 | Houghlines only calculates on the lines scale, checking the whole lines point count(attenuator?) reaching over the threshold.
|
---|
28 |
|
---|
29 | This means houghlinesP can scale better and allow for more accurate results
|
---|
30 |
|
---|
31 |
|
---|
32 | The mathRound does not make difference
|
---|
33 |
|
---|
34 | Drawing forloop with cols, seems to be less effective
|
---|
35 | Even though it does employ same calculations?
|
---|
36 | https://stackoverflow.com/questions/29493267/how-to-detect-lines-using-houghlines-in-opencv-java
|
---|
37 |
|
---|
38 |
|
---|
39 |
|
---|
40 | not enough MP records with more than 10 pages
|
---|
41 |
|
---|