Inquiry about Preprocessing Steps in Azure AI Document Intelligence (Prebuilt Receipt Model)

Question

I am currently working with the Azure AI Document Intelligence service, specifically the Prebuilt Receipt model, and I have a question regarding the preprocessing steps performed automatically by the service.

Could you please clarify the following:

1. Does the Prebuilt Receipt model apply any preprocessing to input images, such as deskewing, denoising, contrast enhancement, or any other transformations?
2. If so, could you provide details on which preprocessing operations are included?

I noticed the 'angle' field in the API response, which seems to indicate some deskewing process, but I would like to confirm the exact scope of preprocessing handled by the service.

Answer 1

Hi 병운 전,

The Azure AI Document Intelligence Prebuilt Receipt model automatically applies several preprocessing steps to input images to optimize text extraction and analysis. These include deskewing, which is confirmed by the presence of the angle field in the API response indicating the rotation applied to correct skew. Additionally, the service performs internal denoising, contrast enhancement, and resolution normalization to improve OCR accuracy, although these operations are not explicitly exposed in the API output. These enhancements are handled automatically by the service, so users do not need to manually apply them before submission. The preprocessing ensures better handling of typical receipt issues such as skewed angles, low contrast, and image noise.

Hope this helps. Do let us know if you have any further queries.

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful.