Text Recognition for Sparse Images via Patch Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional OCR algorithms perform poorly and inaccurately when recognizing text in textually sparse images, often misinterpreting textures or graphical patterns as text due to their design for dense text environments.

Innovation Solution

A text recognition server is configured to analyze images by dividing them into overlapping patches of varying sizes, applying a text detection algorithm to identify text-containing patches, estimating image orientation, and grouping patches into text regions for separate processing by an OCR algorithm, enabling faster and more accurate text recognition in sparse images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional OCR algorithms are applied to textually sparse images, then text recognition is performed across the entire image, but accuracy deteriorates due to misinterpretation of textures or graphical patterns as text

Engineering Contradiction:

Improvetext recognition accuracyVSAvoidfalse positive recognition of non-text as text

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The image is divided into multiple overlapping patches of various sizes. A text detection algorithm is applied to each patch to identify which patches contain text. This segmentation allows the system to focus OCR processing only on text-containing regions, avoiding misinterpretation of non-text areas while maintaining high accuracy in text recognition.

Inventive Principle:

Principle #1Segmentation

2Reliability

If conventional OCR algorithms process the entire image, then comprehensive text coverage is achieved, but processing time increases and efficiency decreases

Engineering Contradiction:

Improvecompleteness of text recognitionVSAvoidprocessing speed

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The image is segmented into overlapping patches that are then grouped into text regions. By applying OCR algorithms only to identified text regions rather than the entire image, the system maintains comprehensive text coverage while significantly reducing processing time and improving efficiency through selective processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

A text detection algorithm is applied to patches before OCR processing to preliminarily identify which regions contain text. This preliminary action filters out non-text areas, allowing the subsequent OCR algorithm to focus only on relevant regions, thereby improving processing speed without sacrificing completeness.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If conventional OCR algorithms analyze the whole image, then all text is potentially detectable, but computational resources are wasted on non-text areas

Engineering Contradiction:

Improvetext detection completenessVSAvoidcomputational energy consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The image is divided into patches and processed selectively. By identifying text-containing patches through a detection algorithm and applying OCR only to those regions, the system ensures all text is detectable while minimizing computational energy consumption by avoiding processing of non-text areas.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Instead of applying OCR to the entire image, the system applies the computationally intensive OCR algorithm only to a subset of patches identified as containing text. This partial action approach maintains text detection completeness while significantly reducing computational energy consumption.

Inventive Principle:

Principle #16Partial or excessive action

4Adaptability or versatility

If conventional OCR algorithms are used on sparse images, then uniform processing is applied, but performance deteriorates due to lack of adaptation to text distribution

Engineering Contradiction:

Improveprocessing adaptabilityVSAvoidtext recognition performance

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The image is segmented into patches of various sizes that are grouped into text regions based on detected text content. This segmentation enables the system to adapt processing to the actual text distribution in the image, applying OCR only where needed and improving recognition performance by matching processing intensity to text density.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts processing based on detected text distribution. By using a text detection algorithm to identify text-containing patches and grouping them into regions, the system adapts its OCR processing to the specific characteristics of each image, improving performance by making processing dynamic rather than uniform.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS9183452B2Text recognition for textually sparse images

Publication Date: 2015.11.10 GOOGLE LLC

AI summary

A text recognition server is configured to recognize text in a sparse text image. Specifically, given an image, the server specifies a plurality of “patches” (blocks of pixels within the image). The system applies a text detection algorithm to the patches to determine a number of the patches that contain text. This application of the text detection algorithm is used both to estimate the orientation of the image and to determine whether the image is textually sparse or textually dense. If the image is determined to be textually sparse, textual patches are identified and grouped into text regions, each of which is then separately processed by an OCR algorithm, and the recognized text for each region is combined into a result for the image as a whole.