Introduction

The manual aims to guide you to get high quality document images, thus ensuring a proper functioning of our systems. Following these recommendations, we guarantee clear and clear images that are crucial for the processing and extraction of precise information.

Ideal Capture Conditions

To ensure best image quality, consider the following environmental and preparatory factors:

Optimal light

• It uses soft and uniform natural or artificial light.
• Avoid shadows and reflections that may obscure parts of the document.
• Put yourself back to a window to take advantage of natural light without generating strong contrasts.

Sense

• Make sure that the camera's lens is clear, dusted and fingerprint free.
• It uses a soft microfibre cloth to clear the lens without scratching it.
• Consider an approach to objects at the bottom.
• Consider the shortterm approach to avoid distortions.

Stability and Nitidity

• Avoid motion of the camera during capture to prevent blurred images.
• Be sure that the document's flat and fixed onto a stable surface.
• Avoid "blur," that's what happens when there's a lot of movement.
• It prevents the opacity of the documents, which do not allow them to be read.
• Avoid using digital filters that modify the captured image.

Suitable Resolution

• Absolute minimum: 1000 pixels on the bigger side of the image. Below this threshold, extraction OCR can return empty fields (CURP, CIC, RFC, names) while the image "looks good" to a clear eye.
• Recommended: 1500 pixels or more on the bigger side for an extraction OCR stable.
• Support forms: JPEG and PNG. Other formats (WebP, HEIC, etc) generate a quality warning and may be misprocessed.

Hardware requirements for an optimum capture

To maximize the quality of the captured images, we recommend using equipment with the following specifications:

House Resolution

• Ideally, use a camera with a resolution of 2 megapixels (MP) or higher.

Camera Sensor

• He preferred type sensors. CMOS (Complementary Metal-Oxide Semiconductor).
• A sensor size of 1 / 2.7 "or greater improves light collection, leading to lighter and less noise images.

Optics

• It uses high quality glass lenses with low distortion.
• An opening of f / 2.0 or less allows more light entry, improving performance under low light conditions.

Image Formatting

• Choose the format RAW for best quality and flexibility in editing, or JPEG (high quality) for a balance between quality and file size.

Camera Setup

• Adjust the ISO between 100 and 400 To minimize noise in the image, especially under low light conditions.

Important considerations about Argentina and Mexico's document

Based on our experience and growing use of our platform by users at Argentina and MexicoWe share some important comments about documents issued in these countries:

Argentina

Documents such as Driver's license and DNI includes visible safety elements, as holograms under some light conditions these holograms can make automatic reading more difficult (OCR) information. scanning the documents under appropriate light conditionsfollowing the suggested guidelines for best results.

Mexico

A number have been historically found above the mean number of Faked documentsMoreover, even among genuine documents, they exist. variations in design, typography and format by region, municipality or issuing authority. OCRespecially if the documents do not follow standardized guidelines.

Properties of the Spanish Alphabet

The use of Spanish alphabetincluding characters with tides and other diacriticsThese elements are usually small and may be more likely to be errors during automatic extraction as compared with the characters of the English alphabet.

Those points are not limitations but factors to be taken into account to better understand why in some cases the reading accuracy may vary.

Additional recommendations

• Capture Software: Using SDK document collection JAAK.
• Evidence: Reconduct tests to check that the configuration and light conditions are optimal.

Quality diagnosis from your integration

If you're integrating against /v4/document/extract and you see empty fields intermittent, you can activate the image quality report to diagnose the problem without waiting to reproduce it in another session.

Leader opt-in X-Include-Quality

By default the response of /v4/document/extract doesn't include quality metrics to maintain light response. If you send the header X-Include-Quality: true, the answer includes a field quality with the result of an analysis of the image received:

{
  "eventId": "...",
  "status": "SUCCESS",
  "content": { "data": { "personal": { ... } } },
  "quality": {
    "width": 608,
    "height": 405,
    "score": 0.32,
    "warnings": ["low_resolution"]
  }
}

Warning code

Automatic cleanup of OCR fields (INE / Voter ID of Mexico)

For model documents VOTER_ID (credential to vote INE) we apply automatic consolidation and validation to the fields extracted before returning them:

• CURP (18 characters) RENAPO. If you do not fulfill the pattern, the field will be returned empty instead of returning a CURP malformed.
• Elector's key (18 alphanumeric characters): valid against the official pattern. OCR (0 vs O, 1 vs I, 5 vs S) are automatically corrected if the character context allows them (letter and digit positions).
• CIC (9 digits of the IDMEX at the bottom of the back): lt's valid to be exactly 9 digits. The IDMEX line stands up to remove it even if it's accompanied by other characters.

If the image is low resolution, the OCR you can read partial characters that don't pass these validations. The visible effect on your integration is an empty field rather than an incorrect value. an empty field to an CURP/ malformed CIC that your system will accept as valid.

Following these recommendations, we guarantee that high quality document photo capturefacilitating the processing and collection of precise information.