Min C. Shin

Does colorspace transformation make any difference on skin detection

Skin detection is an important process in many of computer vision algorithms. It usually is a process that starts at a pixel-level, and that involves a pre-process of colorspace transformation followed by a classification process. A colorspace transformation is assumed to increase separability between skin and non-skin classes, to increase similarity among different skin tones, and to bring a robust performance under varying illumination conditions, without any sound reasonings. In this work, we examine if the colorspace transformation does bring those benefits by measuring four separability measurements on a large dataset of 805 images with different skin tones and illumination. Surprising results indicate that most of the colorspace transformations do not bring the benefits which have been assumed.

Effect of colorspace transformation, the illuminance component, and color modeling on skin detection

Skin detection is an important preliminary process in human motion analysis. It is commonly performed in three steps: transforming the pixel color to a non-RGB colorspace, dropping the illuminance component of skin color, and classifying by modeling the skin color distribution. In this paper, we evaluate the effect of these three steps on the skin detection performance. The importance of this study is a new comprehensive colorspace and color modeling testing methodology that would allow for making the best choices for skin detection. Combinations of nine colorspaces, the presence of the absence of the illuminance component, and the two color modeling approaches are compared. The performance is measured by using a receiver operating characteristic (ROC) curve on a large dataset of 805 images with manual ground truth. The results reveal that (1) colorspace transformations can improve performance in certain instances, (2) the absence of the illuminance component decreases performance, and (3) skin color modeling has a greater impact than colorspace transformation. We found that the best performance was obtained by transforming the pixel color to the SCT or HSI colorspaces, keeping the illuminance component, and modeling the color with the histogram approach.

Objective evaluation of approaches of skin detection using ROC analysis

Skin detection is an important indicator of human presence and actions in many domains, including interaction, interfaces and security. It is commonly performed in three steps: transforming the pixel color to a non-RGB colorspace, dropping the illuminance component of skin color, and classifying by modeling the skin color distribution. In this paper, we evaluate the effect of these three steps on the skin detection performance. The importance of this study is a new comprehensive colorspace and color modeling testing methodology that would allow for making the best choices for skin detection. Combinations of nine colorspaces, the presence or the absence of the illuminance component, and the two color modeling approaches are compared for different settings (indoor or outdoor) and modeling parameters (the histogram size). The performance is measured by using a receiver operating characteristic (ROC) curve on a large dataset of 845 images (consisting more than 18.6 million pixels) with manual ground truth. The results reveal that (1) colorspace transformations can improve performance in certain instances, (2) the absence of the illuminance component decreases performance, and (3) skin color modeling has a greater impact than colorspace transformation. We found that the best performance was obtained on indoor images by transforming the pixel color to the HSI or SCT colorspaces, keeping the illuminance component, and modeling the color with the histogram approach using a larger size distribution.

Tracking Colliding Cells In Vivo Microscopy

Leukocyte motion represents an important component in the innate immune response to infection. Intravital microscopy is a powerful tool as it enables in vivo imaging of leukocyte motion. Under inflammatory conditions, leukocytes may exhibit various motion behaviors, such as flowing, rolling, and adhering. With many leukocytes moving at a wide range of speeds, collisions occur. These collisions result in abrupt changes in the motion and appearance of leukocytes. Manual analysis is tedious, error prone, time consuming, and could introduce technician-related bias. Automatic tracking is also challenging due to the noise inherent in in vivo images and abrupt changes in motion and appearance due to collision. This paper presents a method to automatically track multiple cells undergoing collisions by modeling the appearance and motion for each collision state and testing collision hypotheses of possible transitions between states. The tracking results are demonstrated using in vivo intravital microscopy image sequences. We demonstrate that 1) 71% of colliding cells are correctly tracked; (2) the improvement of the proposed method is enhanced when the duration of collision increases; and (3) given good detection results, the proposed method can correctly track 88% of colliding cells. The method minimizes the tracking failures under collisions and, therefore, allows more robust analysis in the study of leukocyte behaviors responding to inflammatory conditions.

Induction of biphasic changes in perfusion heterogeneity of rat liver after sequential stress in vivo.

Trauma and subsequent sepsis lead to hepatic microcirculation disruption through various molecular mechanisms in which endothelin-1 (ET-1) plays a pivotal role. These stresses are thought to alter hepatic perfusion, heterogeneously leading to a mismatch of oxygen supply and demand. We hypothesize that mild remote stresses prime the liver to sequential sepsis through direct effects on the hepatic lobular flow distribution. We also propose to investigate the extent and the localization of the stress-induced microcirculation disruption. Sprague-Dawley rats were randomly divided into four experimental groups: sham, femur fracture (FFX), cecal ligation and puncture (CLP), and sequential stress (SS). Hepatic intravital microscopy was performed for in vivo assessment of the liver microcirculation flow distribution under baseline and after ET-1 infusion. Red blood cell motion distribution was used to quantify intralobular and interlobular heterogeneity of perfusion (HoP). Intralobular HoP, which reflects lobular regulation sites, was significantly increased in the FFX and CLP groups, but was not changed or decreased in the SS group compared with control. ET-1 infusion exerted opposite effects depending on the pathological condition, further increasing the difference between groups. SS induced decrease in intralobular HoP, contrasted with a significant increase in interlobular HoP, suggesting multiple disruption sites. Our data suggest that increased intralobular HoP may be indicative of a compensatory response to moderate stress; its decrease under sequential stress conditions corresponds with a total breakdown of hepatic lobular flow regulation. This may be another instance of the rich variability characteristic of normal physiology that “decomplexifies” under critical decompensated conditions.

HYDROGEN SULFIDE MODULATES SINUSOIDAL CONSTRICTION AND CONTRIBUTES TO HEPATIC MICORCIRCULATORY DYSFUNCTION DURING ENDOTOXEMIA

Hydrogen sulfide (H₂S) affects vascular resistance; however, its effect on the hepatic microcirculation has not been investigated. Hepatic sinusoidal perfusion is dysregulated during sepsis, contributing to liver injury. Therefore, the present study determined the effect of H₂S on the hepatic microcirculation and the contribution of endogenous H₂S to hepatic microcirculatory dysfunction in an endotoxin model of sepsis. Portal infusion of H₂S increased portal pressure in vivo (6.8 ± 0.2 mmHg before H₂S vs. 8.6 ± 0.8 mmHg peak during H₂S infusion, P < 0.05). Using intravital microscopy, we observed decreased sinusoidal diameter (6.2 ± 0.27 μm before H₂S vs. 5.7 ± 0.3 μm after H₂S, P < 0.05) and increased sinusoidal heterogeneity during H₂S infusion (P < 0.05) and net constriction. Since hepatic H₂S levels are elevated during sepsis, we used the cystathionine γ lyase inhibitor DL-propargylglycine (PAG) to determine the contribution of H₂S to the hypersensitization of the sinusoid to the vasoconstrictor effect of endothelin-1 (ET-1). PAG treatment significantly attenuated the sinusoidal sensitization to ET-1 in endotoxin-treated animals. ET-1 infusion increased portal pressure to 175% of baseline in endotoxemic animals, which was reduced to 143% following PAG treatment (P < 0.05). PAG abrogated the increase in sinusoidal constriction after ET-1 infusion in LPS-treated rats (30.9% reduction in LPS rats vs. 11.6% in PAG/LPS rats, P < 0.05). Moreover, PAG treatment significantly attenuated the increase in NADH fluorescence following ET-1 exposure during endotoxemia (61 grayscale units LPS vs. 21 units in PAG/LPS, P < 0.05), suggesting an improvement in hepatic oxygen availability. This study is the first to demonstrate a vasoconstrictor action of H₂S on the hepatic sinusoid and provides a possible mechanism for the protective effect of PAG treatment during sepsis.

Robust bee tracking with adaptive appearance template and geometry-constrained resampling

Studying and analyzing inter-communication among bees requires tracking of many bees. Manual labeling of bees over many frames is painstaking and time-consuming. Automated tracking is challenging because of the appearance change and unreliable features. This problem is magnified when tracking for a longer period of time is required. We present a method for tracking bees that minimizes the accumulation of error over time by using (1) static and adaptive appearance templates for handling appearance change, and (2) geometry-constrained resampling of particles for handling unreliable features. Evaluation against manually-labeled ground truth demonstrates that our method tracks bees with an RMSE of 8.7 pixels (typical bee length is ≫100 pixels), and 75% position and 58% angular error improvement over a particle filtering based tracking with gaussian modeling of appearance.

Efficient tracking of ants in long video with GPU and interaction

Behavior analysis of social insects requires robust tracking over many frames. Automated tracking methods are not reliable for tracking over long video. And they are prone to a quick accumulation of error from one mis-tracking. However, searching and correcting of mis-tracking is time-consuming. In this paper, we present an efficient method for achieving robust tracking of multiple ants over a long video. First, our method minimizes the user wait time by speeding up tracking with a GPU. Second, it minimizes the amount of data the user needs to validate by automatically searching for potential errors and presenting them for user validation and correction. User studies with three participants on a 10,000 frame video demonstrates that (1) the speed of tracking is 16x faster with GPU optimization, (2) tracking accuracy was 96%, which is a 25% improvement over no user interaction, (3) users examined less than 0.6% of frames for validation and correction.

Liver Microcirculation Analysis by Red Blood Cell Motion Modeling in Intravital Microscopy Images

Intravital microscopy has been used to visualize the microcirculation by imaging fluorescent labeled red blood cells (RBCs). Traditionally, microcirculation has been modeled by computing the mean velocity of a few, randomly selected, manually tracked RBCs. However, this protocol is tedious, time consuming, and subjective with technician related bias. We present a new method for analyzing the microcirculation by modeling the RBC motion through automatic tracking. The tracking of RBCs is challenging as in each image, as many as 200 cells move through a complex network of vessels at a wide range of speeds while deforming in shape. To reliably detect RBCs traveling at a wide range of speeds, a window of temporal template matching is applied. Then, cells appearing in successive frames are corresponded based on the motion behavior constraints in terms of the direction, magnitude, and path. The performance evaluation against a ground truth indicates the detection accuracy up to 84% TP at 6% FP and a correspondence accuracy of 89%. We include an in-depth discussion on comparison of the microcirculation based on motion modeling from the proposed automated method against a mean velocity from manual analysis protocol in terms of precision, objectivity, and sensitivity.

Task-based evaluation of skin detection for communication and perceptual interfaces

Skin detection is frequently used as the first step for the tasks of face and gesture recognition in perceptual interfaces for human-computer interaction and communication. Thus, it is important for the researchers using skin detection to choose the optimal method for their specific task. In this paper, we propose a novel method of measuring the performance of skin detection for a task. We have created an evaluation framework for the task of hand detection and executed this assessment using a large dataset containing 17 million pixels from 225 images taken under various conditions. The parameter set of the skin detection has been trained extensively. Five colorspace transformations with and without the illuminance component coupled with two color modeling approaches have been evaluated. The results indicate that the best performance is achieved by transforming to SCT colorspace, using the illuminance component, and modeling the distribution with the histogram approach. Some conclusions such as the SCT colorspace being one of the best colorspaces are consistent with our previous work, while findings such as the YUV colorspace performing well in this work when it was one of the worst in our previous work are different. This indicates that the performance measured at the pixel-level might not be the ultimate indicator for the performance at the task-level of hand detection. We believe that the users of skin detection will find our task-based results to be more relevant than the traditional pixel-level results. However, we acknowledge that an evaluation is limited by its specific dataset and evaluation protocols.