I have two images of same character (e.g. A) but of different sizes. My question is "what is the feature of both image having same value that means one property of first image is same for the second image. Please reply.

I have two images of same character (e.g. A) but of different sizes. My question is "what is the feature of both image having same value that means one property of first image is same for the second image. Please reply. I try using different properties of regionprops (eccentricity, orientation, extent) and also bware function.

Answer 1

[Image Analyst]

Measurements that are "scale invariant" are those that don't have units of pixels or pixels squared or depend on the actual numbers of pixels being measured, like orientation, Euler number, Solidity, and ratios like circularity = perimeter^2/(4*pi*area).

Measurements that depend on scale are those that have units of pixels, like MajorAxisLength, MinorAxisLength, area, perimeter, EquivDiameter, {PixelList, PixelIdxList, PixelValues, Centroid, etc.

[mamata]

Thanks,I am working for OCR. I am able to extract each character. As per your suggestion the circularity feature is same upto 2 decimal places. please reply me "if the two images are Normalized i.e convert to a single size, then is any information loss", so that recognition is incorrect. If possible then how to normalize.

[Image Analyst]

If an image is shrunk, information is lost. If an image is magnified, information is not necessarily added. Changing the size will not necessarily lose so much information that recognition can no longer be done - it may still be possible to do the same algorithm successfully.