How to find the disparity in image processing.

[ngkamsengpeter]

Given 2 image taken from 2 camera in canonical position, and the baseline is b, we can find the distance Z of an object from the camera. We know that

Z=bf/D where D is disparity.

I read some books but still have no idea of how to find the disparity using c#. Can someone suggest some mathematics and algorithm for me?
I also read about finding the correspondence point of the image but i have no idea of how the corresponding point related to disparity. And what is the unit of disparity we found using algorithm.Is it in unit of pixel and can we plug directly into above equation to find Z in unit of mm or cm? Or how can we convert focal length into pixel?

I also read about the essential matrix and fundamental matrix, do i really need to find them? because if i can find disparity without using these matrix, then i already can find the distance using formula above. This matrices is very difficult to solve.

My supervisor require me to write the entire algorithm by myself without using any library such as opencv. So someone please give me the mathematics behind the algorithm.

 

[karthike]

Hi,
For more details regarding disparity, refer this link
http://siddhantahuja.wordpress.com/category/stereo-vision/

With regard to essential matrix and fundamental matrix, refer to 2nd and 3rd paragraph after proof in the link
http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT10/node3.html

Regards,
karthik

 

[quicknote]

I can understand your confusion and frustration in trying to find the disparity in image processing. Finding the disparity is an important step in depth estimation and is crucial in many applications such as 3D reconstruction and object detection. I will try to provide some guidance and clarification on the concepts you have mentioned.

Firstly, let's understand what disparity means in image processing. Disparity refers to the difference in the location of an object between two images taken from different viewpoints. This difference can be used to estimate the depth of the object from the camera.

To find the disparity, we need to first find the corresponding points in the two images. These points are the same feature or object in both images but appear at different locations due to the difference in viewpoints. This is where the essential matrix and fundamental matrix come into play. These matrices provide a way to find the corresponding points between two images. So, yes, you do need to find these matrices in order to find the disparity.

Now, coming to the mathematics and algorithms, I cannot provide a specific algorithm as it would depend on the specific scenario and the images being used. However, I can provide some general steps that can help you in finding the disparity.

1. Pre-processing: The first step is to pre-process the images to improve their quality and remove any noise. This can include operations like image resizing, noise removal, and image enhancement.

2. Feature detection and matching: The next step is to detect and match features in the two images. This can be done using techniques like SIFT (Scale-Invariant Feature Transform) or SURF (Speeded-Up Robust Features). These algorithms will provide you with the corresponding points between the two images.

3. Calculating the essential matrix: Once you have the corresponding points, you can use them to calculate the essential matrix. This matrix represents the relationship between the two camera viewpoints and is crucial in finding the disparity.

4. Calculating the fundamental matrix: The fundamental matrix is derived from the essential matrix and provides a way to find the disparity between the corresponding points in the two images.

5. Estimating the disparity: Finally, you can use the fundamental matrix to estimate the disparity between the corresponding points. This disparity can then be used in the formula Z = bf/D to find the depth of the object from the camera.

Regarding your question about the unit of disparity, it is usually measured in pixels as it represents the difference in the location of the object in the two images