Extract information from grayscale image

Hi everyone, I need to extract some information from the attached grayscale image. Since this kind of image contains information (multiple lines, shading) I cannot convert to binary image to avoid loss of information. Generally my main objective is to understand the image intensity so that I can identify which one is front or rear dimensrion.

Thank you

Answer 1

[Image Analyst]

Line drawings are ambiguous. You can't tell whether the two upper quadrilaterals are closest to you (so that the "notch" is in the back and you are looking at the underside), or the notch is in the front and you are looking down at the object. A classical line drawing of the 12 sides of a box is similar - you can't tell.

Why do you have to do this anyway? What is the "use case" (to use extreme programmer's lingo)

Answer 2

[missC]

Firstly thanks for your response. Well, seems its complicated. The main objective is to reconstruct the 3D model from the line drawing. Somehow I believe the illustration like shading, line pressure can assist on the recognition. Previously, I converted into line drawing (using morphological operation and thinning algorithm) and currently working on 2D geometrical (junctions,line and regions) entities extraction using chain code. By having only line drawing, I can't recognize which one is front or rear dimension. Could you have any suggestion other than this? Many thanks.

[Image Analyst]

I'm not sure without researching it. It won't be easy. You can start here: http://en.wikipedia.org/wiki/Hidden_line_removal

You didn't answer my second question about why you need to do this. Do you know?

[missC]

As I told you (maybe i did not clarified), it is for 2D line recognition. I converted to binary image and did some preprocessing until get thinned image in order to identify the geometrical entities. Now I am trying to understand, how to program so that by having intensity of image, a machine can understand the bold and pressure part will be the front dimension. In simple manner, for dimension (front and rear) recognition. I am still in the process of this, and everything is for 3D reconstruction