Given the same aspect ratio of images but different resolutions, the Ephesoft system have built in mathematics to automatically deal with differing resolutions of the same image and be able to successfully handle KV extraction.


This feature takes in account the case when a KV extraction rule is configured for a sample of specific resolution, whereas the batch execution input image is of different resolution other than the configured one. Application will internally adjust overlays for the image of Batch Execution by normalizing them.

Instead of storing key-Value coordinates as distance from origin of the image, we would be storing it in normalized form. This way, the coordinates stored would be independent of image size.

For Example, we have following

Key Coordinates: (x, y)

Value Coordinates: (x’, y’)

Image Coordinates: (X, Y)


Normalized Values to be stored would be:

Normalized Key Coordinates: (x/X, y/Y)

Normalized Value Coordinates: (x’/X, y’/Y)

Similarly other values related to KV would be calculated and stored.

During extraction the normalized values would be de-normalized with respect to the input image given for extraction.

Assuming Image Coordinates are: (P, Q)

De-Normalized Key Coordinates: ((x/X)*P, (y/Y)*Q)

De-Normalized Value Coordinates: ((x’/X)*P, (y’/Y)*Q)

These de-normalized Coordinates would be normalized with respect to new image dimensions. Key Value extraction would be performed based on these normalized coordinates.

Was this article helpful to you?


Comments are closed.