Image Comparison - Rotate, Align, and Zoom

Question

I have the following needs. There are several forms, i.e. Spaces - for example, those used in surveys. Those that are not filled with information, I will now call up image templates. Besides image templates, I also have many images, which are essentially image templates filled with information. For example, there is a survey, and there are two spaces to fill in - these are image templates. Many people have filled in the blanks with their personal information, and these are images.

Image templates are scanned in perfect shape. But many scanned images are tilted or misaligned or possibly scaled. Therefore, I have the following requirement - each image must be recognized to which image template it belongs to. After its recognition, it should be correctly skewed, aligned and scaled to the image template.

I know this is a difficult task and therefore I need a library, preferably C# . I found AForge , but so far I have only seen a suitable method for skewing. Essentially, I need a library that accepts an image template and image as input, and sets a flag if the image does not match the image template. But if it matches, it should return the appropriate angle, alignment and scaling.

If you have ideas or use such a library, I will be very grateful.

Wish you all the best,

Peter

Answer 1

The problem is the image registration problem associated with the pattern matching problem.

• image registration

Depending on how the distorted document can be distorted (scale factor, rotation, skew ...), you can register images using a simple hard conversion (i.e. translation + rotation, only two corresponding points are required) to a more complex one, for example, a non-rigid transformation (a more appropriate point is required). Corresponding points can be specified manually, but ideally should be automatically detected.

ITK library includes several image registration methods

• pattern problem

Once your images are aligned, a comparison between the image and a possible template database can be achieved by first extracting the image features and comparing them with your template database. This is very general and needs to be clarified regarding the image used.

There is another way to combine and register images and match patterns.

• Bag function approach, which consists in extracting interesting points from an image (resistant to several types of image deformation), the points generate a signature that characterizes the image, and image comparison is actually a signature comparison.

Answer 2

I worked for a company, Accusoft Pegasus , which has an interesting pattern recognition tool . I did not see their FormFix tool in action after a few years, but it should be able to do what you need.