Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Post-Processing Mask-Based Table Segmentation for Structural Coordinate Extraction

Suren Bandara

TL;DR

The paper tackles the problem of extracting accurate table structure from mask-based segmentation in noisy document images. It introduces a post-processing framework that treats predicted table masks as one-dimensional signals and applies a multi-scale Gaussian convolution with iterative thresholding to reveal explicit row/column coordinates while preserving the original image for OCR. By deriving a midpoint-accumulation density, formulating a variational objective, and implementing an iterative threshold–convolution scheme, the approach yields robust structural boundaries and improves the CASA metric from 67% to 76% on PubLayNet-1M with TableNet and PyTesseract. The method is model-agnostic, resolution-robust, and computationally efficient, providing practical, content-preserving improvements to mask-based table recognition pipelines.

Abstract

Structured data extraction from tables plays a crucial role in document image analysis for scanned documents and digital archives. Although many methods have been proposed to detect table structures and extract cell contents, accurately identifying table segment boundaries (rows and columns) remains challenging, particularly in low-resolution or noisy images. In many real-world scenarios, table data are incomplete or degraded, limiting the adaptability of transformer-based methods to noisy inputs. Mask-based edge detection techniques have shown greater robustness under such conditions, as their sensitivity can be adjusted through threshold tuning; however, existing approaches typically apply masks directly to images, leading to noise sensitivity, resolution loss, or high computational cost. This paper proposes a novel multi-scale signal-processing method for detecting table edges from table masks. Row and column transitions are modeled as one-dimensional signals and processed using Gaussian convolution with progressively increasing variances, followed by statistical thresholding to suppress noise while preserving stable structural edges. Detected signal peaks are mapped back to image coordinates to obtain accurate segment boundaries. Experimental results show that applying the proposed approach to column edge detection improves Cell-Aware Segmentation Accuracy (CASA) a layout-aware metric evaluating both textual correctness and correct cell placement from 67% to 76% on the PubLayNet-1M benchmark when using TableNet with PyTesseract OCR. The method is robust to resolution variations through zero-padding and scaling strategies and produces optimized structured tabular outputs suitable for downstream analysis.

Post-Processing Mask-Based Table Segmentation for Structural Coordinate Extraction

TL;DR

The paper tackles the problem of extracting accurate table structure from mask-based segmentation in noisy document images. It introduces a post-processing framework that treats predicted table masks as one-dimensional signals and applies a multi-scale Gaussian convolution with iterative thresholding to reveal explicit row/column coordinates while preserving the original image for OCR. By deriving a midpoint-accumulation density, formulating a variational objective, and implementing an iterative threshold–convolution scheme, the approach yields robust structural boundaries and improves the CASA metric from 67% to 76% on PubLayNet-1M with TableNet and PyTesseract. The method is model-agnostic, resolution-robust, and computationally efficient, providing practical, content-preserving improvements to mask-based table recognition pipelines.

Abstract

Structured data extraction from tables plays a crucial role in document image analysis for scanned documents and digital archives. Although many methods have been proposed to detect table structures and extract cell contents, accurately identifying table segment boundaries (rows and columns) remains challenging, particularly in low-resolution or noisy images. In many real-world scenarios, table data are incomplete or degraded, limiting the adaptability of transformer-based methods to noisy inputs. Mask-based edge detection techniques have shown greater robustness under such conditions, as their sensitivity can be adjusted through threshold tuning; however, existing approaches typically apply masks directly to images, leading to noise sensitivity, resolution loss, or high computational cost. This paper proposes a novel multi-scale signal-processing method for detecting table edges from table masks. Row and column transitions are modeled as one-dimensional signals and processed using Gaussian convolution with progressively increasing variances, followed by statistical thresholding to suppress noise while preserving stable structural edges. Detected signal peaks are mapped back to image coordinates to obtain accurate segment boundaries. Experimental results show that applying the proposed approach to column edge detection improves Cell-Aware Segmentation Accuracy (CASA) a layout-aware metric evaluating both textual correctness and correct cell placement from 67% to 76% on the PubLayNet-1M benchmark when using TableNet with PyTesseract OCR. The method is robust to resolution variations through zero-padding and scaling strategies and produces optimized structured tabular outputs suitable for downstream analysis.
Paper Structure (23 sections, 3 theorems, 20 equations, 2 figures, 1 algorithm)

This paper contains 23 sections, 3 theorems, 20 equations, 2 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Overview
  5. Derivation of Raw Midpoint Accumulation
  6. Problem Setting and Notation
  7. Variational Objective
  8. Proposed Iterative Threshold–Convolution Scheme
  9. Parameter Selection
  10. Theoretical Analysis
  11. Effect of Gaussian Convolution
  12. Preservation and Variance Increase of Gaussian Mixtures
  13. Role of Thresholding in Mode Selection
  14. Experiments and Results
  15. Experimental Setup
  16. ...and 8 more sections

Key Result

Proposition 1

Let $f_0 \in L^1(\mathbb{R})$ have finite second moment. Convolution of $f_0$ with a Gaussian kernel $\mathcal{G}_\sigma$, produces a smoother distribution with the same mean $\mu = \int x f_0(x)\,dx$ and variance

Figures (2)

  • Figure 1: Example of a noisy column mask output illustrating the challenge of using mask-based segmentation directly for structural coordinate extraction.
  • Figure 2: Evolution of the density $f_n(x)$ during iterative threshold–convolution. Dominant peaks corresponding to column boundaries become increasingly pronounced across iterations.

Theorems & Definitions (6)

  • Proposition 1: Smoothing as Low-Pass Filtering
  • Proof 1
  • Theorem 1: Closure under Gaussian Convolution
  • Proof 2
  • Theorem 2: Nonlinear Mode Suppression
  • Proof 3