I am using estimateAffinePartial2D to match stitch two images together. Sometimes I get a completely wrong result (scaled way too much, rotated way too much, …) as my cameras are handheld, certain movements are impossible (like upside down, …). Is there a way to restrict the possible transformations which estimateAffinePartial2D can make? Is there any other method which might be better suited?

I am using estimateAffinePartial2D to match stitch two images together. Sometimes I get a completely wrong result (scaled way too much, rotated way too much, …) as my cameras are handheld, certain movements are impossible (like upside down, …). Is there a way to restrict the possible transformations which estimateAffinePartial2D can make? Is there any other method which might be better suited?

I am using estimateAffinePartial2D to match stitch two images together. Sometimes I get a completely wrong result (scaled way too much, rotated way too much, …) as my cameras are handheld, certain movements are impossible (like upside down, …). Is there a way to restrict the possible transformations which estimateAffinePartial2D can make? Is there any other method which might be better suited?

My code looks like this:

hope you are doing well today, i am working with a personal project, im using some cables to find the edges, find the center and sample the colors in hsv to then figure out the color and list it in a final image drawn on the frame, yes this is real time.

I have an image of an object I want to put in a setting, and I’ve explored a few different approaches and here are my initial impressions

How we can count triangles in this image:

I have a document in the form of an image. The content of the document is some information in the form of a table. Most often, these are some scanned documents, that is, black and white. However, sometimes there may be originals, where there may be a blue seal, or a company logo of any color (perhaps this information is important for solving my issue).

here the code:

How can the location of the paper rectangle be detected as a rotated in these images using opencv? The big problem is the background which makes it hard to extract only the paper.