Image Spam Detection via Contour Segmentation and Pixel Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting SPAM in images are inefficient due to the ability of SPAM generators to evade signature-based and OCR-based detection techniques, leading to resource-intensive processing and high false positive rates, particularly in real-time email filtering.
Innovation Solution
A system and method that uses a two-color representation of images to identify contours corresponding to symbols, parses text lines into words, and verifies the presence of text by comparing pixel colors, while ignoring noise and variations, and employs Levenshtein distance for similar signature matching to render a SPAM/no SPAM verdict.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If signature-based methods are used to detect SPAM in images, then detection speed is improved, but detection accuracy deteriorates because SPAM generators can easily evade by randomly changing bits
Solution Approach 1:
The patent segments the image into multiple blocks and analyzes each block separately to extract features such as text presence, color distribution, and pattern characteristics. This segmentation approach allows the system to maintain detection speed while improving accuracy by identifying SPAM patterns across multiple regions rather than relying on a single signature comparison.
Solution Approach 2:
The patent replaces the mechanical signature-based comparison method with an automated analysis system that uses algorithms to detect text, analyze color distributions, and identify patterns. This substitution enables the system to automatically adapt to SPAM variations without manual signature updates, improving both speed and reliability.
2Measurement precision
If OCR-based methods are used to detect text in images, then text detection capability is improved, but resource consumption increases dramatically
Solution Approach 1:
The patent applies partial OCR action by first analyzing image blocks for basic characteristics (color distribution, text presence indicators) before applying full OCR only to blocks that show strong text indicators. This selective approach maintains text detection capability while significantly reducing overall resource consumption by avoiding exhaustive OCR on every image block.
Solution Approach 2:
The patent applies different analysis qualities to different regions of the image. Simple statistical analysis is applied to most blocks, while full OCR is applied only to specific blocks that exhibit characteristics indicating text presence. This local quality approach optimizes resource usage by applying computational intensive methods only where necessary.
3Measurement precision
If full OCR is applied to all image blocks, then text detection accuracy is improved, but processing time increases making real-time filtering unacceptable
Solution Approach 1:
The patent implements periodic action by applying a two-stage analysis process: first a quick statistical analysis of all blocks, then full OCR only on blocks that pass the initial filter. This periodic application of intensive processing maintains text detection accuracy while reducing overall processing time to meet real-time filtering requirements.
Solution Approach 2:
The patent performs preliminary analysis of image blocks to identify those containing text before applying full OCR. By pre-screening blocks using faster statistical methods (color distribution, text presence indicators), the system prepares a reduced set of candidates for intensive OCR processing, thereby maintaining accuracy while reducing processing time.
4Reliability
If image filtering is performed on all incoming emails, then SPAM detection capability is improved, but system performance deteriorates due to hardware resource requirements
Solution Approach 1:
The patent segments the email processing workload by dividing images into blocks and analyzing them in parallel. This segmentation enables the system to process multiple image regions simultaneously, maintaining high SPAM detection capability while improving system throughput and performance through efficient resource utilization.
Solution Approach 2:
The patent implements a multi-functional analysis system that performs multiple types of analysis (statistical analysis, text detection, pattern recognition) using a unified framework. This universal approach allows the same system to handle both simple and complex SPAM detection scenarios, improving detection capability while optimizing resource usage across different email types.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system, method and computer program product for identifying spam in an image including identifying a plurality of contours in the image, the contours corresponding to probable symbols (letters, numbers, punctuation signs, etc.); ignoring contours that are too small or too large given the specified limits; identifying text lines in the image, based on the remaining contours; parsing the text lines into words; ignoring words that are too short or too long, from the identified text lines; ignoring text lines that are too short; verifying that the image contains text by comparing a number of pixels of a symbol color within remaining contours to a total number of pixels of the symbol color in the image; and if the image contains a text, rendering a spam/no spam verdict based on comparing a signature of the remaining text against a SPAM template.