Privacy protection in the surveillance video data has received great attention. Although tremendous works have been proposed to provide effective privacy protection techniques, most of the algorithms are based on post-processing that deletes, obscures or encrypts the privacy information after privacy-included raw data are recorded. Consequently, they are vulnerable to raw data leak out, which may lead to unauthorized use. Therefore, it is imperative to develop a new privacy protection scheme which is capable of excluding any privacy information at the video recording phase.

We propose an anonymous camera aiming to protect the privacy of individuals at the video capturing phase by optical masking technique. It effectively reduces the risk of raw data leakage because no privacy information will be recorded by the camera. We implemented a prototype camera, which consists of an infrared (IR) camera, a RGB camera and a liquid crystal on silicon (LCoS) device. We introduce optical design and performance of the anonymous camera, the tracking and masking algorithm as well as the calibration methodology. 

First, the IR mirror reflects infrared light rays into the IR camera, and passes visible rays to the lens. Once an IR image frame is transmitted to the central processing device, the IR face detection algorithm generates the mask pattern where the bare face regions are set to black. The mask is subsequently cast onto the LCoS device, which selectively allows visible rays of unmasked spatial positions to be reflected to the RGB sensor camera. As a result, the RGB camera only captures images with human face masked, thus the person privacy is protected.

Since it is hard to recognize person identities in low-resolution IR images captured from a distance, this system guarantees no recognizable facial images are captured, transmitted, processed, and stored during the entire process.

The prototype of our system is:

The flowchart of the face masking algorithm is as follows:

Note that there exist three image planes in the system: the sensor plane of the IR camera, the charge-coupled device (CCD) and the plane of LCoS device. We assume that these three planes are parallel and precisely aligned pixel by pixel. However, practically, we need to ensure it through image calibration among these images since the prototype is usually not perfectly aligned. We modeled the projective relation among these images by homograpies. The detailed process and figures are included in our paper.

Here are some demo results of our system: