Description:FORM 2
THE PATENTS ACT 1970
(39 of 1970)
&
The Patents Rules, 2003
COMPLETE SPECIFICATION
(See section 10 and rule 13)
1. TITLE OF THE INVENTION: “SYSTEM AND METHOD FOR IRIS DETECTION AND TEMPLATE EXTRACTION IN BIOMETRIC IDENTIFICATION”
2. APPLICANTS:

(A) NAME : MANTRA SOFTECH (INDIA) PRIVATE LIMITED
(B) NATIONALITY : INDIAN
(C) ADDRESS : B-203 SHAPATH HEXA
OPP. GUJARAT HIGH COURT
S. G. HIGHWAY, SOLA
AHMEDABAD 380 060

PROVISIONAL
The following specification describes the invention. þ COMPLETE
The following specification particularly describes the invention and the manner in which it is to be performed.

Field of invention
The present invention relates to system and method for iris detection and template extraction in biometric identification using hierarchical segmentation model and template extraction techniques.
Background of invention
Iris recognition is a biometric identification method that leverages the unique patterns of the human iris to accurately identify individuals. The concept dates back to 1936 when ophthalmologist Frank Burch proposed using iris patterns for identification purposes. This idea was further supported by ophthalmologists Leonard Flom and Aran Safir in the 1980s, who suggested that no two irises are alike, leading to a patent for the iris identification concept in 1987. An implementation of iris recognition was developed in the 1990s, algorithms capable of analyzing iris patterns became the foundation for commercial iris recognition systems.
Early iris recognition systems required individuals to remain stationary and positioned at specific angles close to the camera for accurate image capture. Over the years, iris recognition technology has been integrated into various sectors, including security screening at airports, border control, and national identification programs. For instance, the Government of India enrolled the iris codes of over 1.2 billion citizens in the Unique Identification Authority of India (UIDAI) program to enhance national ID and fraud prevention efforts.
Despite advancements, challenges such as image quality under varying lighting conditions, motion blur, and occlusions continue to affect the accuracy of iris recognition systems. Biometric authentication relies heavily on iris recognition due to its unique and stable physiological characteristics. Traditional methods suffer from limitations in terms of accuracy, processing speed, and adaptability to different environmental conditions.
The present invention overcomes these limitations by employing an advanced hierarchical segmentation model, optimizing the detection and template extraction process while ensuring compliance with international biometric standards.
Object of Invention
The main object of system and method for iris detection and template extraction in biometric identification is to overcome the problem regarding the segmentation techniques, edge detection, and contour analysis for robust iris localization and template extraction.
Yet another object of the present invention is to improve the consistency of biometric templates by introducing a normalization that compensates for variations in iris size and illumination across different datasets.
Further object of the present invention is to ensure that the integration of image quality identification, correction, and hierarchical segmentation techniques results in a novel biometric matching approach that surpasses prior art in accuracy and reliability.
Another object of the present invention is to optimize the method for biometric template extraction by dynamically integrating image quality analysis, adaptive correction, and hierarchical segmentation to provide a highly reliable and efficient iris recognition system.
Another object of the present invention is to implement a hierarchical segmentation models and standardized template extraction techniques for iris detection and biometric template extraction, significantly improving the reliability and efficiency of biometric authentication systems.
These and other objects will be apparent based on the disclosure herein.
Summary of invention
The present invention relates to system and method for iris detection and template extraction in biometric identification. A Data Preparation Framework is designed to convert raw biometric images into a well-labeled dataset through semi-automated annotation techniques. Annotation of distinct components, including sclera, iris, pupil, and background, ensures high-quality input data for training a hierarchical segmentation model. Further, a Hierarchical Segmentation Module is implemented to process background, sclera, iris, and pupil in a step-by-step manner along with deep learning models for enhanced segmentation accuracy and efficiency. Segmentation Masks are Integrated and Processed in Post-processing model for refining sclera masks to improve accuracy and various segmentation outputs are integrated with additional processing layers to enhance final template quality. The extracted biometric template adheres to ISO standards, ensuring interoperability across biometric systems.
The system and method for iris detection and template extraction in biometric applications provides multiple benefits to the end users which are described in the following pages of specification.
Brief description of drawings
Other objects, advantages and novel features of the invention will become apparent from the following detailed description of the present embodiment when taken in conjunction with the accompanying drawings.
Fig. 1 illustrates a block diagram depicting various modules (200) of the iris detection and template extraction system, in accordance with some embodiments of the present disclosure.
Fig. 2 illustrates a flow diagram, depicting a method of operation of the iris detection and template extraction system, in accordance with some embodiments of the present disclosure.
Fig. 3 illustrates input and output images of sclera left_right_process, in accordance with some embodiments of the present disclosure.
Fig. 4 illustrates a sclera mask with top left, bottom left, top right, and bottom right points, in accordance with some embodiments of the present disclosure.
Fig. 5 illustrates a sample sclera mask with accurate curvature, in accordance with some embodiments of the present disclosure.
Fig. 6 illustrates a sample result images.
Detailed Description of Invention
Before explaining the present invention in detail, it is to be understood that the invention is not limited in its application to the details of the construction and arrangement of parts illustrated in the accompany drawings. The invention is capable of other embodiment, as depicted in different figures as described above and of being practiced or carried out in a variety of ways. It is to be understood that the phraseology and terminology employed herein is for the purpose of description and not of limitation.
It is to be also understood that the term "comprises" and grammatical equivalents thereof are used herein to mean that other components, ingredients, steps, etc. are optionally present. For example, an article "comprising" (or "which comprises") components A, B, and C can consist of (i.e., contain only) components A, B, and C, or can contain not only components A, B, and C but also contain one or more other components.
A system and method for iris detection and template extraction for biometric applications employs advanced image processing, computer vision, and deep learning techniques to detect the iris, pupil, and sclera, and extract a distinctive biometric template.
System comprises a Hardware Image Capture Device, preferably a High-resolution camera capable of capturing detailed iris images. Processing Unit comprises a high-performance processor (CPU/GPU) capable of handling deep learning based training models and image processing tasks in real-time.
As shown in fig. 1, iris detection and template extraction system (100)may include, but not limited to, several modules (200) that captures and processes an image of user’s (1) eyes. Said system comprises various modules (200) like a capture module (201), an quality analysis and correction module (203), an iris detection module (205), a template extraction module (207), a hierarchical segmentation module, and a post-processing module (211) for providing various data like a captured data (301), a enhanced image (303), an iris region data (305), standardized biometric templates (307), refined segmented images (309), and final crop and mask images (311) for processing purposes.
As depicted in fig. 2, the capture module (201) is configured to capture (401) user’s (1) eye data preferably in form of image, but also can be configured to capture data in form of video. The captured data (301) of a user’s (1) eye is stored for further processing (401).
The quality analysis and correction module (203) is configured to perform image quality analysis on the captured data (301) by evaluating metrics including, but not limited to, a sharpness, contrast, brightness, noise level, and artifacts. Based on the assessment the quality analysis and correction module (203) performs adaptive corrections on the captured data (301) including, but not limited to, de-noising, contrast enhancement, brightness normalization, and edge sharpening to generate an enhanced image for subsequent processing to acquire an enhanced image (303).
The iris detection module (205) detects and isolates the iris region data (305) within the enhanced image (303) by implementing advanced image processing, computer vision techniques, and deep learning-based semantic segmentation, thereby minimizing false positives even under suboptimal conditions. Said Module contains pre-processing image enhancement techniques like edge detection, contour analysis, and feature extraction models to improve the quality of the iris region data (305).
The template extraction module (207) generates a compact and distinctive iris template by extracting the iris, pupil, and sclera regions from the detected iris region data (305) to acquire standardized biometric templates (307). Further normalization is preformed on standardized biometric templates (307)for variations in size, illumination, and other inconsistencies generating the standardized biometric templates (307) that comply with ISO standards for biometric scoring mechanisms. The final standardized biometric templates (307) are encoded into a format suitable for efficient storage and comparison.
Data preparation framework performs labeling and dataset creation for further processing. The system processes raw image data into a labeled dataset using semi-annotated techniques. Each component (iris, pupil, sclera, background) is labeled. These labels are used to train the hierarchical segmentation models for segmentation. The labeled data follows a hierarchical structure, aiding in segmentation by clearly defining regions of interest (ROI).
The hierarchical segmentation module (209) is configured and trained for implementing a step-by-step detection and segmentation process for segmenting the background, iris, pupil, and sclera regions from the detected standardized biometric templates (307) to acquire refined segmented images (309). The image quality analysis and correction module (203) is integrated with said the segmentation process for dynamically optimizing both the input and output of the hierarchical segmentation module (209). The hierarchical segmentation module (209) is a multi-stage segmentation module, which is developed with separate phases for detecting the background, sclera, iris, and pupil. In Step-by-Step Segmentation, Each step in the segmentation process targets a specific object (e.g., iris, sclera) to ensure better accuracy and efficiency. The hierarchical segmentation model enhances processed image using deep learning and machine learning techniques to improve precision in detection.
The hierarchical segmentation module (209) initiates processing and normalizing an input data for further processing. First stage of the segmentation comprises sclera segmentation by utilizing multi‑scale region embedding and edge‑aware refinement for acquiring refined sclera segment. Second stage comprises iris segmentation by utilizing convolutional neural network (CNN) and convolutional block attention models to focus on iris texture, iris boundaries and avoid occlusions. Third stage comprises pupil segmentation based on CNN to learn a precise, circular pupil shape in the dark central region to ensure pixel accuracy and perfect circularity. Forth stage comprises boundary refinement by implementing an active contours and morphological cleanup. All the segmented images or masks undergoes the normalization and quality assessment and flags low‑quality images for manual review or reacquisition.
The post-processing module (211) processes the refined segmented images (309) to create the final crop and mask image (311). Said module utilizes five images to generate the final crop and mask (311) with accurate curvature including iris, pupil, original image, sclera full mask [mask from model], and sclera generated mask [polygon mask] with the correct curvature. Post-processing steps includes a sclera mask post-processing based on left-right point saturation and integration of acquired masks. Sclera mask post-processing refines and enhances the sclera mask for higher accuracy. initial step of processing the sclera mask includes a sclera_left_right_saturation process function which plays a crucial role in addressing the issue of in-congruence between the left and right sclera masks. A novel technique named as “left-right point saturation” is implemented to address gaps in the sclera mask. As shown in fig. 3, the input image illustration reveals a slight gap on the left side, through the implementation of the processing methodology, said module effectively rectify this gap, leading to a notable enhancement in the ensuing image. The sclera_left_right_process includes sequence of operations that are performed on refined segmented images (309) starting form converting an image to grayscale, fixing left-right gaps, smoothing using Gaussian blur, dilating and eroding, and apply binary threshold to obtain the final binary sclera mask from the refined segmented images (309).
As shown in fig. 4, the post-processing module (211) deploys contour detection to locate the four corner points of the sclera mask. Said module iterates through a list of contours, which represent distinct objects or regions within an image. Subsequently, the two smallest and two largest points are selected.
A mathematical approach of Bézier curves is utilized to model the curvature of the sclera and accurately detect points in the image. The system calculates the top and bottom sclera points using Bézier curve equations. Quadratic Bézier Curves are employed for modeling the precise curvature of the sclera to improve detection accuracy. As the curve is completely contained in the convex hull of its control points, the points can be graphically displayed and used to manipulate the curve intuitively. Affine transformations such as translation and rotation can be applied on the curve by applying the respective transform on the control points of the curve which are four corner points of the sclera mask in present disclosure. Calsulated results from equation helps to accurately represent the curvature of the sclera region. Said four points collectively define the corners of the sclera. For the subsequent steps, Arcs are created individually for the top two points and the bottom two points. By drawing these arcs, said module establishes a foundation for generating two distinct masks: one for the upper section and another for the lower section. These masks are eventually merged to form a unified result, stored as a variable combined_binary_mask. The sclera mask with accurate curvature is generated as illustrated in fig. 5.
Post-Processing Refinement involves Bitwise “OR” Operations and black pixel count calculations to fine-tune the masks for refining the sclera mask, ensuring accurate representation of sclera boundaries. Adjustments are made to predefined points for better alignment and accuracy in final sclera detection.
Furthermore, In the Iris post-processing stage, the post-processing module (211) utilize iris and sclera masks generated by the hierarchical segmentation model, and crop the original image based on the dimensions of the respective iris or sclera mask. The Post-processing module (211) employs two techniques, namely blending and blurring or feathering the edges of objects within an image. Firstly, the iris mask undergoes dilation to expand its edges. Subsequently, the post-processing module (211) apply blur or feathering by imposing a Gaussian blur to the edges of the iris mask image. The Gaussian blur is imposed on the image, which identifies contours, and blends the original and blurred images based on these contours, creating an output image. Said Gaussian blur process involves convolution, threshold, and contour identification. A generated result has the dimensions of the iris mask. Three images are selected for further processing: the iris mask, the generated result, and the original cropped image. Blending is applied to incorporate only the iris region.
The sclera mask undergoes a bitwise “NOT” operation. An inverted copy of the sclera mask is created and blurring or feathering is applied using Gaussian blur to the edges of the inverted sclera mask. Said process results in three images: the blended iris image, the sclera mask, and the sclera mask focusing only on areas with lowest and highest pixels. These three images serve as inputs for the blending technique.
In the final step, the upper and lower eyelid portions are blended onto the previously obtained iris. The resulting image represents the culmination of the entire process, featuring a final crop and mask image (311) with the sclera portion
After each mask (sclera, iris, and pupil) has been processed, they are combined using advanced integration methods. Additional post-processing is performed for Mask Combination to optimize and refine the masks to generate a final extracted template. The post-processing module (211) creates the final crop and mask image (311) with the correct curvature. Said module is specifically configured to generate images as per the standards to ensure precision in the final output images by making suitable adjustment. The image quality analysis and correction module (203) is integrated with a feedback loop that adjusts image capture parameters including exposure, focus, and sensor gain based on historical quality data, thus reducing the frequency of suboptimal image acquisition at the source.
The template extraction module (207) features a normalization sub-module that compensates for variations in iris size and illumination, ensuring that the resulting biometric template maintains consistency across diverse datasets. Hence, system may also be configured to generate an iris template for specific subject conditions, like elderly individual, non-optimal gaze towards system and limited availability for computing input data. The extracted template is optimized for matching such scenarios and generates an iris template from a subject exhibiting non-frontal or off-angle gaze, the extracted template is optimized for matching an iris template under conditions of limited computational resources. Said formulation emphasizes that the method is tailored for specific subject conditions (elderly, non-optimal gaze, limited computing availability) and ensures that the resulting templates are suitable for effective matching in biometric identification.
The present invention has beneficial advantages that it utilizes a comprehensive approach to enhance the accuracy and efficiency of biometric identification processes. The system's utilization of advanced image processing, hierarchical segmentation models and adaptive post-processing methods ensures reliable and secure authentication in various applications such as access control, identity verification, and surveillance systems.
The invention has been explained in relation to specific embodiment. It is inferred that the foregoing description is only illustrative of the present invention and it is not intended that the invention be limited or restrictive thereto. Many other specific embodiments of the present invention will be apparent to one skilled in the art from the foregoing disclosure.
All substitution, alterations and modification of the present invention which come within the scope of the following claims are to which the present invention is readily susceptible without departing from the invention. The scope of the invention should therefore be determined not with reference to the above description but should be determined with reference to appended claims along with full scope of equivalents to which such claims are entitled.

List of Reference Numerals
1 User
100 Iris Detection and Template Extraction System
200 Modules
201 Capture Module
203 Quality Analysis and Correction Module
205 Iris Detection Module
207 Template Extraction Module
209 Hierarchical segmentation Module
211 Post-Processing Module
300 Data
301 Captured Data
303 Enhanced Image
305 Iris Region Data
307 Standardized Biometric Templates
309 Refined Segmented Images
311 Final Crop and Mask Image
, Claims:We Claim:
1. An iris detection and template extraction system (100), comprising:
an capture module (201) configured to capture data in form of image or video and provide captured data (301);
a quality analysis and correction module (203) configured to perform image quality analysis and adaptive correction on the captured data (301) through evaluating metrics such as sharpness, contrast, brightness, noise level, and artifacts to acquire an enhanced image (303);
an iris detection module (205) configured to detect and isolate an iris region from the enhanced image (305) to acquire iris region data (305);
a template extraction module (207) configured to extract the iris, pupil, and sclera regions from the detected iris region data (305) to acquire standardized biometric templates (307);
an hierarchical segmentation module (209) configured for segmenting the iris, pupil, and sclera regions from the standardized biometric templates (307) to acquire refined segmented images (309);
a post-processing module (211) configured to process the refined segmented images (309) to create the final crop and mask image (311) with the correct curvature;
characterized in that, the quality analysis and correction module (203) is integrated with a hierarchical segmentation module (209) for dynamically optimizing both the input and output of step-by-step segmentation process while each step in the segmentation process targets a specific object (e.g., iris, pupil, sclera);
the post-processing module (211) processes five images of the refined segmented images (309) including iris, pupil, original image, sclera full mask, and sclera generated mask with the correct curvature to create the final crop and mask image (311) with accurate curvature, followed by a left-right-point saturation process for sclera mask detection and processing to effectively rectify gaps while addressing the issue of in-congruence between the left and right sclera masks to enhance the quality of the resulting image.
2. The iris detection and template extraction system (100) as claimed in claim 1, wherein a data preparation framework for labeling raw image data into a labeled dataset using semi-annotated techniques to label each component like iris, pupil, sclera, and background to train the hierarchical segmentation models embedded hierarchical segmentation module (209) for segmentation.
3. The iris detection and template extraction system (100) as claimed in claim 1, wherein a contour detection in post-processing module (211) to locate four corner points of the sclera mask and implementing Quadratic Bézier Curves for modeling the precise curvature of the sclera, in which two separate arcs are created individually for the top two points and the bottom two points, generating two distinct masks, one for the upper section and another for the lower section that are eventually merged to form a unified result to get sclera mask with accurate curvature improving detection accuracy.
4. The iris detection and template extraction system (100) as claimed in claim 1, wherein the quality analysis and correction module (203) is further integrated with a feedback loop that adjusts image capture parameters including exposure, focus, and sensor gain based on historical quality data, thus reducing the frequency of suboptimal image acquisition at the source.
5. The iris detection and template extraction system (100) as claimed claim 1, wherein the template extraction module (207) further comprises a normalization sub-module that compensates for variations in iris size and illumination, ensuring that the resulting biometric template maintains consistency across diverse datasets.
6. A method for an iris detection and template extraction, comprising the steps of:
capturing data in form of image or video using a capture module (201) and stored as captured data (301) for further processing;
performing image quality analysis and correction on the captured data (301) by evaluating metrics such as sharpness, contrast, brightness, and noise level using a quality analysis and correction module (203) to acquire an enhanced image (303);
detecting and isolating the iris region from the enhanced image (305) using an iris detection module (205) to acquire iris region data (305);
extracting the iris, pupil, and sclera regions from the detected iris region data (305) using a template extraction module (207) to acquire standardized biometric templates (307);
segmenting the iris, pupil, and sclera regions from the standardized biometric templates (307) using a hierarchical segmentation module (209) to acquire refined segmented images (309); and
post-processing the refined segmented images (309) to create the final crop and mask image (311) with the correct curvature using post-processing module (211);
wherein the quality analysis and correction module (203) is integrated with the hierarchical segmentation module (209) for dynamically optimizing both the input and output of step-by-step segmentation process while each step in the segmentation process targets a specific object (e.g., iris, pupil, sclera) to ensure better accuracy and efficiency, the post-processing module (211) processes five images of the refined segmented images (309) including iris, pupil, original image, sclera full mask, and sclera generated mask with the correct curvature to create the final crop and mask image (311) with accurate curvature, followed by a left-right-point saturation process for sclera mask detection and processing to effectively rectify gaps while addressing the issue of in-congruence between the left and right sclera masks to enhance the quality of the Refined Segmented Images (309).
7. The method for an iris detection and template extraction as claimed in claim 6, wherein the step of dynamically adjusting image capture parameters such as exposure, focus, and sensor gain based on historical image quality data obtained from the quality analysis and correction module (203), thereby optimizing the input for subsequent processing.
8. The method as claimed in claim 6, wherein the adaptive image correction is implemented through the quality analysis and correction module (203) in real-time through the use of a deep neural network that predicts optimal correction parameters, enhancing the overall quality of the image prior to iris detection.
9. The method as claimed in claim 6, wherein the iris detection module (205) utilizes advanced semantic segmentation techniques combined with edge detection and contour analysis to robustly detect the iris region even under suboptimal imaging conditions to generate iris region data (305).
10. The method as claimed in claim 6, wherein the standardized biometric templates (307) are optimized for specific scenarios and generates an iris templates from a user (1) exhibiting non-frontal or off-angle gaze, and an iris templates under conditions of limited computational resources, making the system adaptive for such specific subject conditions and ensures that the resulting templates are suitable for effective matching in biometric identification.
Dated this on 30th May, 2025.