Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
133 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
46 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Shadow Detection: A Survey and Comparative Evaluation of Recent Methods (1304.1233v1)

Published 4 Apr 2013 in cs.CV and cs.RO

Abstract: This paper presents a survey and a comparative evaluation of recent techniques for moving cast shadow detection. We identify shadow removal as a critical step for improving object detection and tracking. The survey covers methods published during the last decade, and places them in a feature-based taxonomy comprised of four categories: chromacity, physical, geometry and textures. A selection of prominent methods across the categories is compared in terms of quantitative performance measures (shadow detection and discrimination rates, colour desaturation) as well as qualitative observations. Furthermore, we propose the use of tracking performance as an unbiased approach for determining the practical usefulness of shadow detection methods. The evaluation indicates that all shadow detection approaches make different contributions and all have individual strength and weaknesses. Out of the selected methods, the geometry-based technique has strict assumptions and is not generalisable to various environments, but it is a straightforward choice when the objects of interest are easy to model and their shadows have different orientation. The chromacity based method is the fastest to implement and run, but it is sensitive to noise and less effective in low saturated scenes. The physical method improves upon the accuracy of the chromacity method by adapting to local shadow models, but fails when the spectral properties of the objects are similar to that of the background. The small-region texture based method is especially robust for pixels whose neighbourhood is textured, but may take longer to implement and is the most computationally expensive. The large-region texture based method produces the most accurate results, but has a significant computational load due to its multiple processing steps.

Citations (344)

Summary

  • The paper demonstrates that large-region texture-based methods yield superior detection accuracy by effectively differentiating shadows from objects.
  • It categorizes shadow detection approaches into chromacity, physical, geometry, and texture methods to systematically evaluate their robustness under varied conditions.
  • Enhanced shadow removal is shown to significantly improve object tracking performance, underscoring the importance of appropriate method selection in video analysis.

Shadow Detection: A Survey and Comparative Evaluation of Recent Methods

This paper provides a comprehensive survey and comparative evaluation of methods for detecting moving cast shadows, a critical task for enhancing object detection and tracking in video sequences. The surveyed methods are categorized into a feature-based taxonomy comprising four primary categories: chromacity, physical, geometry, and texture-based approaches. By systematically evaluating these approaches on a series of performance metrics, the authors aim to illuminate the strengths and limitations inherent in each.

The geometry-based method is easy to implement when objects have distinct and predictable forms, allowing for straightforward shadow modeling. Despite its potential robustness, its assumptions about object shape and shadow direction limit its applicability across diverse environments. In scenarios where object forms and orientations are well-defined, particularly pedestrian-centric scenes, this method performs adequately.

Chromacity-based approaches, leveraging color information, provide a swift implementation and operation workflow. These methods assume that shadows darken a region while preserving color, operating under the principle of color constancy. However, the efficacy of chromacity-based approaches diminishes in the presence of noise and low-saturation scenes, indicating a strong trade-off between computational efficiency and environmental sensitivity.

The physical methods improve on simple chromacity models by accommodating local shadow behavior adaptation. By using advanced color models that account for variations in illumination, these approaches can yield more accurate results in complex lighting conditions. Nonetheless, they struggle in scenarios where object and background spectral properties are closely aligned.

Texture-based methods perform robustly across various illumination conditions by assessing texture consistency between frames and backgrounds. This category is divided into small-region and large-region approaches, with texture correlation being their primary tool. The small-region methods, while computationally expensive, excel in textured backgrounds, whereas large-region methods, such as those employing gradient-based correlation, offer superior accuracy by integrating color with texture features.

Quantitative evaluations reveal that the large-region texture-based method outperforms its peers in terms of detection and discrimination rates across diverse experimental conditions, including varying levels of color saturation. This method utilizes larger candidate regions to better leverage textual features and discriminate between shadows and objects effectively.

Practically, shadow detection improvements correlate with enhanced tracking accuracy. By testing shadow removal’s impact on various tracking algorithms, the paper demonstrates that optimized shadow detection leads to appreciable gains in tracking performance, emphasizing the importance of robust shadow removal processes in vision-based surveillance systems.

In sum, shadow detection remains a nuanced task with methodological selection often contingent on the specific application context. Future research directions include integrating diverse features to create composite models that inherit the benefits of multiple approaches, thereby ameliorating application-specific limitations. Such innovations, alongside advancements in computational efficiency, could further refine shadow detection and enhance the reliability of computer vision systems.