Serban Oprisescu

Measurements with ToF Cameras and Their Necessary Corrections

The most important characteristic of time-of-flight (ToF) cameras is the ability to measure the distance to each image pixel. Thus, for each pixel, information on both its amplitude and distance to the camera are available. However, technological problems inherent to the acquisition principle lead to inaccuracies in estimating both characteristics: on one hand, there are errors in estimating the distance, especially for far-distance pixels. On the other hand, the detected amplitude decreases with the distance. Part of these inaccuracies are corrected with special camera-calibration software. In this paper, we propose two methods that attempt to further correct each information based on the other one. First, the amplitude image is enhanced by using distance information: a pixel-wise, distance-based correction of the amplitude brings to light details otherwise unnoticeable. Secondly, an amplitude-based distance modification corrects some of the distance estimation errors for far-distance pixels.

Action recognition using time of flight cameras

The paper deals with the issue of action recognition as an application of the new 3D time-of-flight (ToF) camera, exploiting the special ability of the device to measure distances. Segmentation of moving people is straightforward from the distance information and subsequent steps of the processing chain follow in a classical way. We describe the first results on action recognition using ToF camera distance images for a simple task of deciding actions of a single person. The total variation of a function reveals as a very useful feature in these applications.

Analysis of Pap smear images with iso- and edge-contours

A novel method for the detection and segmentation of nuclei and cells in Pap smear images is introduced. The method is based on a geometric analysis of iso- and edge-contours. For nuclei detection we employ isocontours taken at different levels of intensity and we report best detection (object) recall values as well as best segmentation precision values. For cell outline detection, we employ traditional edge-based contours and show that with a simple radial analysis one can detect the outline even in agglomerated cells - which so far has been approached rather hesitantly. The system was tested on three different databases.

Automatic pap smear nuclei detection using mean-shift and region growing

The Babes-Papanicolaou test (also known as Pap smear) is a method of cervical cancer screening used to detect abnormal cells which are or can become cancerous. Since the visual inspection of pap smears is very time consuming, the need for automatic methods is required. This paper presents an algorithm for the automatic detection of nuclei within pap smears images. The algorithm relies in the highly effective mean-shift filtering method which enhances the contrast of nuclei areas. The segmentation consists of a region growing with starting points taken from the image gradient map. Size and eccentricity measures are used to keep only nuclei from the segmented regions. The method is validated on two different pap smear test databases and the detection rate is above 91%.

Automatic segmentation of infantile hemangiomas

Infantile hemangiomas are the most common types of tumors that are found in infants and have an incidence of approximately 10% in the common population. Although most infantile hemangiomas are self-involuting, due to their fast proliferation they may threaten vital anatomical structures and physiological functions; also, the involution process may take up to several years. An accurate monitoring of the progress of hemangioma growth and regression is essential. We thus suggest using a computer aided follow-up monitoring of these lesions by an automatic detection and quantifying the lesion dynamic: regression or proliferation. In this study we propose some image enhancements methods and also a preliminary color based segmentation. We have tested our methods on 25 hemangioma cases and compared the automated segmentation results with clinician-determined segmentation using an area percentage error.

A new contour extraction algorithm for ToF images

This paper presents a new contour extraction algorithm for time-of-flight (ToF) camera distance images. Experimental results and comparisons with the classical grey level contour extractors are presented. A mathematical model for contour validation is developed, and finally, an edge alignment method is proposed.

Automatic evaluation of hemangiomas for follow-up monitoring

Infantile hemangiomas are the most common types of tumors with an incidence of approximately 10% in the common population. An accurate monitoring of the progress of hemangioma growth and regression is essential for an effective treatment. This study presents an automatic evaluation of the evolution of hemangioma on a follow-up series of images based on color and area features. A color constancy approach is applied to correct the variation of ambient illumination. The hemangioma segmentation uses a two-level thresholding approach with some post-processing methods. The proposed method has been tested on 25 hemangioma cases annotated by clinicians.

Histogram and motion based intrusion detection and tracking algorithms for ToF cameras

The paper deals with the issue of intrusion detection in video surveillance using the time-of-flight (ToF) camera, exploiting the special ability of the device to measure distances. Two different detection methods (based on histogram and motion estimation) are presented and compared. Finally, a simple tracking algorithm is proposed and extended to automatically track of multiple moving persons.

Automatic measurement of infantile hemangiomas

Infantile hemangiomas (IH) are a type of benign vascular tumors that appear within the first 5 months of life. The assessment of lesion size and its evolution in time is done manually by the physician, using a ruler, and this measurement is not very accurate. This paper presents a method for automatic measurement of the IH size. The work is divided in two parts: automatic computation of the size of one millimeter in pixels, based on the Hough transform and the total variation, and automatic segmentation based on K-means clustering and a 2D total variation filtered image. The segmentation performance was evaluated on 20 IH images and a mean border error of 13.56% was obtained.

Automatic segmentation of infantile hemangiomas within an optimally chosen color space

Infantile hemangiomas (IH) are benign vascular tumors, most of them appearing in the first weeks and developing until six months of age. The evaluation of the lesion size is usually made by the physician through manual measurement, which is inaccurate. This paper presents an algorithm for the automatic segmentation of the hemangioma region, relying on the Maximum a Posteriori (MAP) classification method. The segmentation result is improved by regularization with discrete Markov fields (MAP-Markov). Then, a further improvement is performed, by eliminated distant non-hemangioma pixels. The optimal color space is chosen before segmentation, from five different color spaces, by iteratively computing the segmentation error 10 times on each color space and each of the 40 images from the database. The segmentation performance is evaluated in terms of border error.