Security applications of image processing difficulties and processor selection (1)

Introduction

The quality of life as people on the requirements and the major trends of the global fight against terrorism, as well as digital technology itself continues to progress, relying on fingerprint recognition, iris recognition, face recognition biometric technology programmes and video monitoring programmes is becoming increasing the personal, family, business and social security of an important means.

Biometric programme involves four steps: image capture, image preprocessing, features sampling, match analysis; and video surveillance programmes mainly including image acquisition and image processing, image processing and transmission, image display and image management. It is not difficult to see, whether it is a biometric or video surveillance, image preprocessing is required. In fact, image preprocessing algorithm complexity, flexibility, image-processing chips resources, as well as the processing time will run directly on the entire system produces a decisive influence. Therefore, image preprocessing for the entire security programmes is a difficult and crucial tasks that directly determine the subsequent image processing and analysis of accuracy and convenience.

Image processing and analysis

According to the purpose of image preprocessing can be divided into clear image on the acquisition, image recognition, pretreatment and image pre-processing before compression.

Which, on the acquisition and image processing mainly includes clarifications on CMOS or CCD image sensor photosensitive cells of inconsistent for subsequent correction, on the physical environment and sensor acquisition of image a differential compensation (such as backlight), as well as on acquisition to the original image denoising. Although the difficulty of preprocessing algorithm itself is not great, but as the popularity of the real needs, especially in the PIX is large, this algorithm or DSP processing power to high demands.

And image recognition pre-backup the purpose, you may need to destroy the original pixel and distribution, so that subsequent for feature extraction.

This preprocessing algorithm of difficulty depending on the identification of occasions. To be integrated later in recognition algorithm section, select the appropriate DSP. Image compression pre-backup mainly refers to YUV422 becomes YUV420, RGB to YUV, etc. This class handles often have real-time requirements, if a software implementation, the handling performance have higher requirements; if you use a hardware implementation, although performance is guaranteed, but the hardware costs will rise.

At the same time, according to the application, image preprocessing can be divided into biometric applications in image processing and video surveillance applications in image processing.

For the biometric application to fingerprint identification, for example, the pretreatment include fingerprint image enhancement, fingerprint image binarization, fingerprint image thinning, fingerprint image thinning post-processing. And video monitoring applications in image processing mainly refers to the image sensor output continuous image analysis, gets enough information, and through the auto white balance and gamma (Gamma) correction, auto focus, auto exposure, backlight compensation to improve the image of the actual effects.

Image processing challenges

Both biometric or video surveillance, its image preprocessing is facing the following challenges: first, the user requirements on image quality, image preprocessing algorithm complexity, image preprocessing main chip processing power and storage space to a more demanding requirements; second, the user on the image of real-time processing and transmission, on the one hand, increasingly demanding requirements of image preprocessing algorithm to optimize and streamline the main chip for image pre-processing of kernel processing power, internal bus architecture, data transfer capability, peripheral interface, as well as the overall framework and the hardware instruction set support for preprocessing algorithm set higher requirements; third, different to the image and video coding/decoding algorithm the algorithm with unified industry standards and a clear evolution roadmaps, image preprocessing algorithm not only there is no uniform standards and clear direction, even for the most part, programme providers it is through these "secrets" of personalized algorithm as a competitive weapon.

In addition, as the different application fields, demand and the evolution of technology itself, the original algorithms continuously upgrade, the new algorithm will emerge that require image processing chip has a higher flexibility and adaptability. Fourth, for programme providers, not only reflected competitive core algorithm needs to prevent the unauthorized reading, or copy, and whether the biometric or video surveillance, its image data often involving privacy, hence the need to provide security assurance can be trusted. The above two aspects that require image processing chip must provide a reliable and complete processing platform.

[1] [2] [3] [4]