Detecting ROI based on edges(Vertical) Density in image

I am working on Extraction of number plate Just by find "Rectangular Areas" And I want Analyze the density of "Edges" in CAR image. . . and finalizing the Region of Interest. . . But don't know where to start. . . PLZ can anybody Help me... ?????

Answer 1

[David Young]

If you have the Image Processing Toolbox, look at the edge function as a start for finding edges. It may be important to set a good scale parameter, so you should look at the Canny method.

If you need to find "vertical" edges (presumably those oriented close to the direction of columns in the image array) you may want to design your own edge detector. This is usually a combination of smoothing and differencing (using conv2 and fspecial) and thresholding.

For finding the density of edges, consider smoothing the edge output by convolution (see conv2) with a smoothing filter (see fspecial).

Then consider thresholding to find areas of high edge density, perhaps applying some morphological operations (bwmorph) to simplify the structure, and choosing regions the right sort of size and shape (using regionprops).

How to start turning this into code? Well, there are many examples in the demos that come with the Image Processing Toolbox. Work through them and you'll find lots of ideas for how you can tackle your problem.