Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clicks2Line: Using Lines for Interactive Image Segmentation

Chaewon Lee, Chang-Su Kim

TL;DR

The paper addresses interactive image segmentation by reducing user clicks for elongated objects through an adaptive input strategy that can switch between clicks and lines. It introduces Clicks2Line, which uses a line-generation pipeline with a line candidate map $\\boldsymbol{\mathcal{L}}$, a target map $\\mathcal{T}$, and a weight map $\\mathcal{W}$ to select an optimal line via $i^{*}=\\arg \\max_{i} (\\mathcal{L} \\times \\mathcal{W})^{i}$ and then converts the line intersection with $\\mathcal{T}$ into two input clicks. The method is validated on GrabCut and Berkeley datasets with a ViT-B backbone and NoC metrics, showing competitive or superior performance in most scenarios, especially for elongated shapes, while qualitatively reducing the number of required user interactions. This work demonstrates that line-guided guidance can significantly lower user effort in interactive segmentation and suggests broader applicability of line-based hints for long objects.

Abstract

For click-based interactive segmentation methods, reducing the number of clicks required to obtain a desired segmentation result is essential. Although recent click-based methods yield decent segmentation results, we observe that substantial amount of clicks are required to segment elongated regions. To reduce the amount of user-effort required, we propose using lines instead of clicks for such cases. In this paper, an interactive segmentation algorithm which adaptively adopts either clicks or lines as input is proposed. Experimental results demonstrate that using lines can generate better segmentation results than clicks for several cases.

Clicks2Line: Using Lines for Interactive Image Segmentation

TL;DR

The paper addresses interactive image segmentation by reducing user clicks for elongated objects through an adaptive input strategy that can switch between clicks and lines. It introduces Clicks2Line, which uses a line-generation pipeline with a line candidate map , a target map , and a weight map to select an optimal line via and then converts the line intersection with into two input clicks. The method is validated on GrabCut and Berkeley datasets with a ViT-B backbone and NoC metrics, showing competitive or superior performance in most scenarios, especially for elongated shapes, while qualitatively reducing the number of required user interactions. This work demonstrates that line-guided guidance can significantly lower user effort in interactive segmentation and suggests broader applicability of line-based hints for long objects.

Abstract

For click-based interactive segmentation methods, reducing the number of clicks required to obtain a desired segmentation result is essential. Although recent click-based methods yield decent segmentation results, we observe that substantial amount of clicks are required to segment elongated regions. To reduce the amount of user-effort required, we propose using lines instead of clicks for such cases. In this paper, an interactive segmentation algorithm which adaptively adopts either clicks or lines as input is proposed. Experimental results demonstrate that using lines can generate better segmentation results than clicks for several cases.
Paper Structure (8 sections, 1 equation, 2 figures, 1 table)

This paper contains 8 sections, 1 equation, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Proposed Algorithm
  3. Input selection: click vs. line
  4. Line generation
  5. Experiments
  6. Implementation Details
  7. Performance Assessment
  8. Conclusions

Figures (2)

  • Figure 1: Comparison of existing algorithm and the proposed algorithm. Row (a) shows the input needed to segment the target region for the SimpleClickliu2023simpleclick method, while row (b) shows that of the proposed method. The line generation process of the proposed method is also illustrated below.
  • Figure 2: Qualitative comparison of algorithms